Merge latest upstream Long Videos

README.md

@@ -35,20 +35,23 @@
 
 - `Dumas H3 Long Videos`
   - Inputs/outputs: the current upstream `MiniMax-H3-Longvideos` sampler surface, exposed under the existing `DumasH3LongVideos` key for saved Dumas workflows.
-  - The local Dumas prompt-engineering fork has been removed from this node. Long Videos now wraps the upstream sampler/engine directly so it can track the source project again.
+  - The local Dumas prompt-engineering fork has been removed from this node. Long Videos now wraps the upstream sampler/engine/runtime/audio/conditioning/shot-plan modules directly so it can track the source project again.
   - Upstream compatibility keys `H3LongVideos`, `H3LongVideosFL2VA`, `H3LongVideosV1`, and `H3LongVideosREF2VA` are also registered to the same class.
   - The old Dumas browser widget grouping script is disabled for this node because it targeted controls that no longer exist on the upstream sampler.
+  - `handoff_frames` extends the upstream last-frame handoff: `1` keeps the current single keyframe behavior; higher values keep that final-frame keyframe and add earlier tail frames from the previous shot as claimed reference context for the next beat.
   - Upstream license text is included in [`H3_LONGVIDEOS_UPSTREAM_LICENSE.txt`](./H3_LONGVIDEOS_UPSTREAM_LICENSE.txt).
 
 - `Dumas H3 Latent Upscale Params`
   - Inputs: `mode`, `model_name`, `method`, `width`, `height`, `device`, `precision`, `sampler_name`, `scheduler`, `steps`, `denoise`, `megapixels`, `tile_width`, `tile_height`, `overlap`, `fade_width`, `fade_height`, `overlap_mode`, `overlap_blend`, `tile_size_mode`, `grid_rows`, `grid_cols`, `spatial_w_overlap`, `spatial_h_overlap`, `min_tile_size`, `masked_area_noise`, `brightness_match`, `dynamic_fade`, `dynamic_fade_min`, `chunk_length`, `temporal_overlap`, `resize_conditioning`, `anchor_strength`
   - Output: `latent_upscale_param`
-  - Bundles the optional latent-space upscaler settings used by `Dumas H3 Long Videos` before decode, so the main node can rebuild conditioning at the target size and run a short refinement pass with your chosen sampler, scheduler, step count, denoise, and the full upstream spatial split controls.
+  - Legacy helper from the abandoned Dumas Long Videos fork. The current upstream-backed `Dumas H3 Long Videos` node does not consume this socket; it uses the upstream latent-upscale controls on the Long Videos node itself.
 
 - `Dumas H3 Beat Prompt`
   - Inputs: authored through the custom front-end beat editor
   - Output: `prompt`
-  - Builds one H3 prompt block per beat, with quick controls for per-shot timing, continuity, ref behavior, anchor additions, soundscape, and music while staying compatible with direct text editing.
+  - Builds an upstream-compatible Long Videos prompt: optional scene paragraph, optional character sheet, then one blank-line-separated textbox per beat.
+  - Per-beat helpers only emit upstream-supported state directives: `remove:` / `removed:` / `off:` and `add:` / `wear:` / `wearing:`.
+  - Old Dumas-only beat directives such as `seconds:`, `continuity:`, `ref_mode:`, `ref_noise_aug:`, `anchor_add:`, `soundscape:`, and `music:` are stripped from the generated prompt so they are not sent to the upstream node as visible text.
 
 - `Dumas H3 Prompt Curator`
   - Inputs: `action_prompt`, `anatomy_guard`, `subject_count_guard`, optional `anchor`, optional `soundscape`, optional `bgm`, optional `ref_1` through `ref_9`

dumas_h3_beat_prompt.py

@@ -2,11 +2,46 @@ import json
 
 
 _DEFAULT_BEAT = "Describe this beat."
-_DEFAULT_STATE = {"beats": [{"text": _DEFAULT_BEAT}]}
+_DEFAULT_STATE = {"scene": "", "character_sheet": "", "beats": [{"text": _DEFAULT_BEAT}]}
+_LEGACY_DIRECTIVE_PREFIXES = (
+    "seconds",
+    "duration",
+    "continuity",
+    "ref_mode",
+    "ref_noise_aug",
+    "anchor_add",
+    "overall_soundscape",
+    "soundscape",
+    "non_diegetic_music",
+    "music",
+    "wardrobe",
+    "enter",
+    "exit",
+)
 
 
 def _clone_default_state():
-    return {"beats": [{"text": _DEFAULT_BEAT}]}
+    return {
+        "scene": "",
+        "character_sheet": "",
+        "beats": [{"text": _DEFAULT_BEAT}],
+    }
+
+
+def _strip_legacy_directives(text):
+    """Remove directives from the abandoned Dumas Long Videos fork.
+
+    The upstream Long Videos node sends unknown field labels to the model as text,
+    so this builder strips the old managed controls rather than emitting prompts
+    that ask H3 to draw labels such as "seconds:" or "music:" in the frame.
+    """
+    kept = []
+    for line in str(text or "").splitlines():
+        lowered = line.strip().lower()
+        if any(lowered.startswith(f"{name}:") for name in _LEGACY_DIRECTIVE_PREFIXES):
+            continue
+        kept.append(line)
+    return "\n".join(kept).strip()
 
 
 def _parse_beat_prompt_state(value):
@@ -21,6 +56,8 @@ def _parse_beat_prompt_state(value):
         except Exception:
             return _clone_default_state()
 
+    scene = str(raw.get("scene") or "")
+    character_sheet = str(raw.get("character_sheet") or "")
     beats = []
     for item in list(raw.get("beats") or []):
         if isinstance(item, dict):
@@ -30,15 +67,25 @@ def _parse_beat_prompt_state(value):
         beats.append({"text": text})
 
     if not beats:
-        return _clone_default_state()
-    return {"beats": beats}
+        beats = [{"text": _DEFAULT_BEAT}]
+    return {
+        "scene": scene,
+        "character_sheet": character_sheet,
+        "beats": beats,
+    }
 
 
 def _assemble_beat_prompt(state):
     parsed = _parse_beat_prompt_state(state)
     chunks = []
+    scene = str(parsed.get("scene") or "").strip()
+    if scene:
+        chunks.append(scene)
+    character_sheet = str(parsed.get("character_sheet") or "").strip()
+    if character_sheet:
+        chunks.append(character_sheet)
     for beat in parsed["beats"]:
-        text = str(beat.get("text") or "").strip()
+        text = _strip_legacy_directives(beat.get("text") or "")
         if text:
             chunks.append(text)
     return "\n\n".join(chunks)
@@ -46,9 +93,9 @@ def _assemble_beat_prompt(state):
 
 class DumasH3BeatPromptNode:
     DESCRIPTION = (
-        "Build a MiniMax H3 prompt from one textbox per beat, with a front-end beat "
-        "editor that can append directive examples and expose per-shot controls for "
-        "timing, continuity, ref behavior, anchor additions, soundscape, and music."
+        "Build an upstream MiniMax H3 Long Videos prompt: optional scene paragraph, "
+        "optional character sheet, then one blank-line-separated textbox per beat. "
+        "Per-beat helpers only emit directives the upstream node understands."
     )
     RETURN_TYPES = ("STRING",)
     RETURN_NAMES = ("prompt",)

dumas_h3_longvideos_audio.py

@@ -0,0 +1,319 @@
+# H3-LongVideos -- https://github.com/Smite79/MiniMax-H3-LongVideos
+# Copyright (c) 2026 Smite79. All rights reserved.
+# Redistribution, in whole or in part, requires written permission.
+# This notice may not be removed or altered. See LICENSE.
+"""Audio policy shared by conditioning and soundtrack assembly."""
+
+from dataclasses import dataclass
+
+import torch
+import comfy.nested_tensor
+from h3_runtime import temporal_shape
+
+
+@dataclass(frozen=True)
+class ShotAudio:
+    speech: bool
+    sounded: bool
+    voiced_only: bool
+    silence_enabled: bool
+    lead_seconds: float
+    latent_fps: int
+    # The tail. Everything after the line's expected end is pinned the way the lead
+    # pins everything before its start. All three default off, so a ShotAudio built
+    # the old way -- six positional arguments -- behaves exactly the old way.
+    line_seconds: float = 0.0      # planner's estimate of the spoken line
+    tail_seconds: float = 0.0      # free audio kept after that estimate; 0 = no tail pin
+    frame_count: int = 0           # the shot in pixel frames; the audio T comes from it
+
+    @property
+    def pinned(self):
+        return self.silence_enabled and not self.speech and not self.sounded
+
+    @property
+    def lead_frames(self):
+        if not self.speech or self.lead_seconds <= 0:
+            return 0
+        return round(self.lead_seconds * self.latent_fps)
+
+    @property
+    def tail_frames(self):
+        """Audio latent frames pinned at the END of a dialogue shot.
+
+        The lead pins the opening so the line cannot start early; nothing pinned the
+        close, and a 2s line in a 9s shot left 7s of open branch in a shot the model
+        knows has a voice in it -- which is where speech carries on past the line, or
+        doubles it. The free span is lead + the line's estimate + tail_seconds; the
+        rest is held at encoded silence. The model chooses WHEN to speak, so the
+        margin is the author's dial: a clipped word costs more than a second of babble.
+        Off unless the shot speaks, the margin is set, and at least half a second would
+        be pinned -- a sliver is not worth the risk of clipping."""
+        if (not self.speech or self.tail_seconds <= 0 or self.line_seconds <= 0
+                or self.frame_count <= 0):
+            return 0
+        total = temporal_shape(self.frame_count)[2]
+        free = self.lead_frames + round((self.line_seconds + self.tail_seconds) * self.latent_fps)
+        tail = total - free
+        return tail if tail >= round(0.5 * self.latent_fps) else 0
+
+
+_SILENT_UNIT = {"lat": None, "key": None}
+
+# ---------------------------------------------------------------------------
+# NOTHING HERE IS SYNTHESISED ANY MORE. Removed on the report: "Just get rid of
+# the ambient sounds all together. They sound horrid. Go back to the model's
+# natural audio."
+#
+# What was here built the soundtrack's non-vocal half out of shaped noise: a room
+# tone from the scene's own wording (synth_ambient, over a table of recipes, with
+# plain_bed under it as a floor) and 21 foley recipes laid into the shots whose
+# audio branch is pinned to silence (foley_for, over _hits/_band/_room and later
+# _contact/_flow/_creak, timed off the picture's own movement for footsteps).
+#
+# It went in because a shot pinned to silence cannot get audio from the model at
+# all -- prompt text never opens a branch -- so auto_sound was writing sounds into
+# prompts that could not make them. That reasoning was sound and the thing it built
+# still did not pass: reported first as footsteps sounding like heartbeats and a
+# bathroom that tapped, and then, once both of those measured clean, as horrid
+# anyway. Synthesis that measures right and sounds wrong is the end of that road.
+#
+# So the audio is the model's, whole. H3 is a joint model and the audio branch is
+# where its sound comes from; the prompt still describes what a shot sounds like,
+# which is the half that was always doing the real work.
+#
+# The consequence, which is real and is reported in info rather than left to be
+# discovered: a shot with no line and no sound you wrote is pinned to silence and
+# is now SILENT. The pin is not a bug and is deliberately untouched -- it is what
+# stops a free branch filling itself with babble and a face lip-syncing to it.
+# Write the sound into the beat to open the branch on purpose, or wire a recording
+# to ambient_audio, which is played under the finished track and conditions
+# nothing. mix_ambient below is that path, and it is all that is left here.
+# ---------------------------------------------------------------------------
+
+
+def _seamless_loop(x, n, sr):
+    """[C, M] -> [C, n], looped with a crossfade so the join does not click.
+
+    Plain tiling puts a discontinuity at every repeat, once per loop length. In a
+    bed that is meant to sit under everything unnoticed, a regular click is the one
+    thing that gets noticed -- the same objection that made the silence latent
+    ping-pong its interior rather than tile it. Here the material is real audio
+    being PLAYED rather than a latent being conditioned on, so it cannot be
+    reversed: a room tone read backwards is fine, but footsteps are not. Crossfade
+    instead, which works on both."""
+    m = int(x.shape[-1])
+    if m <= 0:
+        return None
+    if m >= n:
+        return x[..., :n]
+    fade = min(int(0.25 * sr), m // 4)
+    if fade < 1:
+        reps = -(-n // m)
+        return x.repeat(1, reps)[..., :n]
+    # OVERLAP-ADD the tail onto the head, and shorten the unit by the overlap. The
+    # unit then runs x[m-fade] .. x[m-fade-1], so tiling it steps between samples
+    # that were adjacent in the source and there is no discontinuity anywhere.
+    #
+    # Measured, because the obvious construction is wrong: appending the crossfade
+    # to the END of a full-length unit leaves it finishing on x[fade-1] while the
+    # next repeat starts on x[0], which are not adjacent -- a 2s tone that does not
+    # divide evenly gave a 64x jump at the join, worse than plain tiling's 41x.
+    t = torch.linspace(0.0, 1.0, fade, dtype=x.dtype, device=x.device)
+    head = x[..., :fade] * t + x[..., m - fade:] * (1.0 - t)
+    unit = torch.cat([head, x[..., fade:m - fade]], dim=-1)
+    if int(unit.shape[-1]) < 1:
+        reps = -(-n // m)
+        return x.repeat(1, reps)[..., :n]
+    reps = -(-n // int(unit.shape[-1]))
+    return unit.repeat(1, reps)[..., :n]
+
+
+def mix_ambient(audio, sr, bed, level):
+    """Lay an ambient bed UNDER a finished soundtrack. -> (waveform, note).
+
+    The bed is PLAYED, not conditioned on: it is the file, at the level asked for,
+    under whatever the model generated. That is the whole reason to do it here
+    rather than in the sampler -- ambience needs no cooperation from a joint model,
+    has nothing to lip-sync to, and so cannot put a voice in a wordless shot. The
+    conditioning path can only steer the branch toward something bed-LIKE, and on a
+    shot with a line it competes with the line.
+
+    Defensive throughout, like the silence latent: any failure returns the audio
+    untouched with a note saying so, because a bed is a nicety and a render is not.
+    """
+    try:
+        if audio is None or bed is None or float(level or 0.0) <= 0.0:
+            return audio, ""
+        w = bed.get("waveform") if isinstance(bed, dict) else None
+        if w is None or not int(getattr(w, "ndim", 0)):
+            return audio, ("ambient_audio is wired but carries no waveform, so nothing "
+                           "was laid under the soundtrack")
+        w = w[0] if w.dim() == 3 else w              # [B, C, M] -> [C, M]
+        if w.dim() != 2 or w.shape[-1] < 2:
+            return audio, ("ambient_audio is too short to loop, so nothing was laid "
+                           "under the soundtrack")
+        w = w.detach().to(dtype=audio.dtype, device=audio.device)
+        b_sr = int((bed.get("sample_rate") if isinstance(bed, dict) else 0) or 0)
+        # RESAMPLE, or the bed plays at the wrong speed and pitch. Linear is coarse
+        # for music and inaudible on a room tone, which is what this input is for.
+        resampled = ""
+        if b_sr > 0 and b_sr != int(sr):
+            want = max(2, int(round(w.shape[-1] * float(sr) / float(b_sr))))
+            w = torch.nn.functional.interpolate(
+                w.unsqueeze(0), size=want, mode="linear", align_corners=False)[0]
+            resampled = f", resampled from {b_sr} Hz"
+        ch = int(audio.shape[1])
+        if int(w.shape[0]) != ch:
+            w = (w.mean(dim=0, keepdim=True).repeat(ch, 1) if int(w.shape[0]) > ch
+                 else w[:1].repeat(ch, 1))
+        n = int(audio.shape[-1])
+        loop = _seamless_loop(w, n, int(sr))
+        if loop is None:
+            return audio, ""
+        out = audio + loop.unsqueeze(0) * float(level)
+        # NORMALISE rather than clip. Clipping a bed that pushed a loud line over
+        # the top distorts the LINE, which is the thing worth keeping.
+        peak = float(out.abs().max())
+        gain = ""
+        if peak > 1.0:
+            out = out / peak
+            gain = f", and the mix was scaled by {1.0 / peak:.2f} to stop it clipping"
+        secs = w.shape[-1] / float(sr)
+        return out, (f"an ambient bed was laid under the whole soundtrack at level "
+                     f"{float(level):.2f} -- {secs:.1f}s of audio{resampled}, looped "
+                     f"with a crossfade so the join does not click{gain}. It is your "
+                     f"file, played under what the model generated: it conditions "
+                     f"nothing, so it cannot put a voice in a wordless shot the way "
+                     f"an inferred bed did. Shots pinned to silence keep their silent "
+                     f"conditioning and get the bed on top, which is what makes a "
+                     f"wordless shot sound like a room instead of a mute")
+    except Exception as exc:
+        return audio, (f"the ambient bed could not be mixed ({type(exc).__name__}), so "
+                       f"the soundtrack is unchanged")
+
+
+_SILENCE_STATUS = {"asked": 0, "applied": 0, "why": ""}
+
+
+_SILENT_SECONDS = 2
+
+
+_SILENT_EDGE = 4
+
+
+def _silent_audio_latent(audio_vae, frame_count, fps):
+    """A keyframe audio latent of actual SILENCE, or None if it cannot be made.
+
+    H3 is a JOINT model: the mouth follows the audio branch. On a shot with no
+    scripted line the branch is otherwise unconditioned, and an unconditioned audio
+    branch invents a voice -- which the picture then lip-syncs to. The lips-closed
+    sentence is arguing with a stream that has already decided someone is talking.
+
+    REBUILT 2026-09-05, from measurements against the real VAE rather than from
+    reasoning. The previous version encoded one second, kept a SINGLE interior
+    frame and repeated it, on the argument that silence is homogeneous. It is not,
+    in latent space: encoded silence has genuine frame-to-frame variation (delta
+    mean 0.002-0.004, max 0.021), and a repeated frame has a delta of exactly
+    0.000000. That is a flat signal no encoder produces, and a model handed
+    conditioning outside its own distribution has every reason to disregard it --
+    which is an audio branch back to inventing a voice, with the report saying
+    silence went on.
+
+    The fix that version was avoiding is real too: tiling the whole encoded second
+    end to end leaves a 25x spike at each join (0.554 against 0.022), once per
+    second, which is a metronome in the conditioning of a joint model.
+
+    So: encode two seconds, drop the padded ends, and PING-PONG the interior --
+    forward, reversed, forward. Every join repeats a frame, so there is no seam,
+    and the interior statistics are the encoder's own. Measured over a 9s shot:
+
+        one frame repeated   peak 0.000686   delta mean 0.000000   max 0.000000
+        whole 2s tiled       peak 0.000314   delta mean 0.017451   max 0.554715
+        interior ping-pong   peak 0.000566   delta mean 0.002039   max 0.021159
+
+    where the encoder's own interior is mean 0.0021, max 0.0212. Decoded peak
+    0.000566 on a +/-1.0 scale is about -65 dBFS: silence.
+
+    Everything here stays defensive. Shapes are CHECKED against what the layout
+    expects rather than assumed, and any failure returns None so the shot falls
+    back to an unconditioned branch instead of breaking the render -- the caller
+    reports when that happens, so it is no longer a silent failure.
+    """
+    try:
+        sr = int(getattr(audio_vae, "audio_sample_rate", 0) or 0)
+        if sr <= 0:
+            return None
+        _, _, want_t = temporal_shape(frame_count, fps)
+        if want_t <= 0:
+            return None
+        key = (id(audio_vae), sr)
+        block = _SILENT_UNIT.get("lat") if _SILENT_UNIT.get("key") == key else None
+        if block is None:
+            # CHANNELS LAST. comfy.sd.VAE.encode() does `pixel_samples.movedim(-1, 1)`
+            # before handing off, so the audio VAE -- which wants [B, 2, L] -- must be
+            # given [B, L, 2]. Passing [B, 2, L] raises inside the encoder, and an
+            # early version did exactly that: swallowed by the guard below, so the
+            # whole layer silently did nothing.
+            #
+            # Two seconds, encoded ONCE and cached. Encoding a full 15s shot instead
+            # cost a VAE pass big enough to OOM mid-render on a 16GB card, where the
+            # failure again degraded silently to no conditioning at all.
+            enc = audio_vae.encode(torch.zeros((1, sr * _SILENT_SECONDS, 2)))
+            if enc is None or enc.dim() != 4 or enc.shape[1] != 32:
+                return None
+            if enc.shape[-1] <= 2 * _SILENT_EDGE + 1:
+                return None
+            block = enc[..., _SILENT_EDGE:-_SILENT_EDGE].detach().to("cpu").clone()
+            _SILENT_UNIT["lat"] = block
+            _SILENT_UNIT["key"] = key
+        n = block.shape[-1]
+        if n < 1:
+            return None
+        # Forward, reversed, forward... Each join repeats a frame, so the seam that
+        # plain tiling leaves is gone while the interior variation is the encoder's.
+        pieces, have, i = [], 0, 0
+        while have < want_t:
+            piece = block if i % 2 == 0 else torch.flip(block, dims=[-1])
+            pieces.append(piece)
+            have += n
+            i += 1
+        out = torch.cat(pieces, dim=-1)[..., :want_t].clone()
+        if out.shape[-1] != want_t:
+            return None
+        return out
+    except Exception:
+        return None                         # never fail a render for a nicety
+
+
+def _pin_audio_silence(latent, silence, lead_frames=None, tail_frames=0):
+    """Start target audio at encoded silence and preserve the requested span(s).
+
+    lead_frames None pins the whole shot. Otherwise the first lead_frames and the
+    last tail_frames are held at silence and the span between is left to the model
+    -- that is where the line goes. The tail is clipped to what the lead leaves, so
+    the two can never overlap. Nothing pinned at all is a no-op, reported as False
+    so the caller does not count it as applied."""
+    try:
+        video, audio = latent["samples"].unbind()
+        silence = silence.to(device=audio.device, dtype=audio.dtype)
+        if silence.shape != audio.shape:
+            return False
+        audio_mask = torch.ones_like(audio[:, :1])
+        if lead_frames is None:
+            audio_mask.zero_()
+        else:
+            t = audio.shape[-1]
+            n = min(t, max(0, int(lead_frames)))
+            m = min(t - n, max(0, int(tail_frames or 0)))
+            if n <= 0 and m <= 0:
+                return False
+            if n > 0:
+                audio_mask[..., :n] = 0
+            if m > 0:
+                audio_mask[..., t - m:] = 0
+        latent["samples"] = comfy.nested_tensor.NestedTensor((video, silence))
+        latent["noise_mask"] = comfy.nested_tensor.NestedTensor(
+            (torch.ones_like(video[:, :1]), audio_mask))
+        return True
+    except Exception:
+        return False

dumas_h3_longvideos_conditioning.py

@@ -0,0 +1,310 @@
+# H3-LongVideos -- https://github.com/Smite79/MiniMax-H3-LongVideos
+# Copyright (c) 2026 Smite79. All rights reserved.
+# Redistribution, in whole or in part, requires written permission.
+# This notice may not be removed or altered. See LICENSE.
+"""Decisions about which pictures may condition a shot."""
+
+import torch
+import node_helpers
+from h3_runtime import (H3_FPS, AUDIO_LATENT_FPS, _empty_av_latent, _resize, ref_image_canvas,
+                        frame_levels)
+from h3_audio import _SILENCE_STATUS, _silent_audio_latent, _pin_audio_silence
+
+
+def may_carry_room(previous_cast, current_cast, tagged_names):
+    """A previous frame is safe as a reference only when it adds no subject."""
+    previous = [name for name in (previous_cast or ()) if name]
+    current = set(current_cast or ())
+    tagged = set(tagged_names or ())
+    return bool(previous) and all(name in current for name in previous) \
+        and not any(name in tagged for name in previous)
+
+
+def may_carry_frame(previous_cast, current_cast, tagged_names):
+    """A previous frame of the SAME room is safe as a reference claimed with everyone in it.
+
+    Unlike may_carry_room, somebody this shot does not describe may be in it: the claim
+    names them, and they are still in that room. Refused only for an empty frame, or
+    one holding somebody whose own portrait also rides this shot -- two pictures of one
+    person is how a second one gets drawn."""
+    previous = [name for name in (previous_cast or ()) if name]
+    current = set(current_cast or ())
+    tagged = set(tagged_names or ())
+    return bool(previous) and not any(name in tagged and name in current
+                                      for name in previous)
+
+
+def recoverable_subject(cast, tagged_names, returning_names, captured):
+    """Return the sole safe recovered subject, or an empty string."""
+    people = [name for name in (cast or ()) if name]
+    if len(people) != 1:
+        return ""
+    name = people[0]
+    return name if (name not in set(tagged_names or ())
+                    and name in set(returning_names or ())
+                    and captured.get(name) is not None) else ""
+
+
+KEYFRAME_SAFE_AUG = 0.99       # below this, a ref aug would soften the keyframe too
+
+# What ONE boundary is allowed to claim it measured. Wider than any real per-pass drift,
+# narrow enough that a bad frame -- a flash, a cut to black, a frame the model lost --
+# cannot swing the estimate. The median across boundaries does the real rejecting.
+LEVEL_GAIN_CAP = 0.12          # in log-gain, so +-12.7% of contrast
+LEVEL_OFFSET_CAP = 0.05
+# The within-shot term is believed only when boundaries AGREE on its sign, and even then
+# only this far: within-shot change is often the author's (a light switched off), so it is
+# the half of the signal that cannot be trusted on its own.
+LEVEL_SHOT_GAIN_CAP = 0.015
+LEVEL_SHOT_OFFSET_CAP = 0.010
+LEVEL_AGREE = 2.0 / 3.0
+LEVEL_MIN_OBS = 3
+# What the correction may do to one handoff, whatever it measured. A cut should not carry
+# a visible grade step: shot N's last frame reaches the video uncorrected while N+1 is
+# sampled from a corrected keyframe, so an uncapped correction trades burn-in for a pop at
+# every join -- the same class of complaint, differently shaped.
+LEVEL_GAIN_LO, LEVEL_GAIN_HI = 0.80, 1.25
+LEVEL_OFFSET_BOUND = 0.02
+# Below this a frame is too flat for a contrast RATIO to mean anything.
+LEVEL_MIN_SIGMA = 0.01
+
+
+class HandoffLevels:
+    """Takes the grade the chain adds to itself back out of the handoff.
+
+    THE MEASUREMENT, which is the whole reason this needs no scene list. At every
+    boundary the render holds two pictures that are SUPPOSED to be the same frame: K,
+    the handoff it gave the shot, and R, frame one of what came back -- the model's own
+    reproduction of K, from a keyframe labelled sigma 0.001. Nothing was asked to change
+    between them, so everything separating them is the chain's own doing and none of it
+    is the author's intent. That is the one difference in the loop that can be corrected
+    without guessing at anybody's lighting, and R costs nothing to look at: it is the
+    frame trim_seam throws away.
+
+    A beat that walks into a darker room moves K, and R follows it there. So the level is
+    never anchored, never compared to shot 1, and never compared to a target -- only K
+    against its own reproduction, boundary by boundary.
+
+    WHAT IT WILL NOT FIX. Clipping already baked into earlier shots, because the VAE
+    clamps every decode and headroom spent is gone. Softening, which is a different
+    measurement and a different cause. Anything spatial -- ghosting, local burn, identity
+    drift. A tone curve with a knee in it, since this is affine per channel; the residual
+    in the report is how that would show itself. The first boundary, which has nothing to
+    measure yet. And a deliberate monotone move -- a film that dims every single beat --
+    loses a bounded, reported fraction of itself."""
+
+    def __init__(self):
+        self._bg, self._bo = [], []      # per boundary: K -> R, the chain's own drift
+        self._sg, self._so = [], []      # per shot: R -> last frame, believed only on agreement
+        self.applied = []                # (gain, offset) actually used, for the report
+
+    def observe(self, given, repro, last=None, pre_up_last=None):
+        """Record one boundary. given is the keyframe this shot got, repro is frame one
+        of what it produced, last is its final frame, pre_up_last the handoff it hands on.
+
+        last/pre_up_last are how the pre-upscale handoff and the post-upscale output are
+        put in the same frame of reference: their difference IS the pipeline's own offset,
+        measured on one frame that went through both, so it can be subtracted from the
+        K->R reading instead of being mistaken for drift. With latent_upscale off they are
+        the same frame and the term is zero."""
+        gm, gs = frame_levels(given)
+        rm, rs = frame_levels(repro)
+        if gm is None or rm is None:
+            return False
+        if float(gs.min()) < LEVEL_MIN_SIGMA or float(rs.min()) < LEVEL_MIN_SIGMA:
+            return False
+        ug = torch.zeros(3)
+        uo = torch.zeros(3)
+        lm, ls = frame_levels(last) if last is not None else (None, None)
+        if pre_up_last is not None and lm is not None:
+            pm, ps = frame_levels(pre_up_last)
+            if pm is not None and float(ps.min()) >= LEVEL_MIN_SIGMA:
+                ug = torch.log(ls / ps)
+                uo = lm - pm
+        self._bg.append((torch.log(rs / gs) - ug).clamp(-LEVEL_GAIN_CAP, LEVEL_GAIN_CAP))
+        self._bo.append((rm - gm - uo).clamp(-LEVEL_OFFSET_CAP, LEVEL_OFFSET_CAP))
+        if lm is not None and float(ls.min()) >= LEVEL_MIN_SIGMA:
+            self._sg.append(torch.log(ls / rs))
+            self._so.append(lm - rm)
+        return True
+
+    def _agreed(self, rows, cap):
+        """The median of rows, but only per channel where at least LEVEL_AGREE of them
+        share its sign. A within-shot change the boundaries disagree about is content, not
+        drift, and content must not be corrected."""
+        out = torch.zeros(3)
+        if len(rows) < LEVEL_MIN_OBS:
+            return out
+        st = torch.stack(rows)
+        med = st.median(dim=0).values
+        agree = ((st * med.sign().unsqueeze(0)) > 0).float().mean(dim=0)
+        keep = agree >= LEVEL_AGREE
+        return torch.where(keep, med.clamp(-cap, cap), out)
+
+    def estimate(self):
+        """(gain_log, offset) the chain is drifting by per boundary, per channel."""
+        if not self._bg:
+            return None, None
+        g = torch.stack(self._bg).median(dim=0).values + self._agreed(self._sg, LEVEL_SHOT_GAIN_CAP)
+        o = torch.stack(self._bo).median(dim=0).values + self._agreed(self._so, LEVEL_SHOT_OFFSET_CAP)
+        return g, o
+
+    def gains(self, strength):
+        """(gain, offset) as 3-vectors, or (None, None) when there is nothing worth doing.
+
+        Separate from note() because more than one frame leaves a shot -- the handoff,
+        and any face captured for a return several shots later -- and they have to carry
+        the SAME grade. A recovered face arriving at a different exposure from the shot
+        around it would be a new bug of exactly the kind this is fixing."""
+        g, o = self.estimate()
+        if g is None or strength <= 0:
+            return None, None
+        gain = torch.exp(-float(strength) * g).clamp(LEVEL_GAIN_LO, LEVEL_GAIN_HI)
+        off = (-float(strength) * o).clamp(-LEVEL_OFFSET_BOUND, LEVEL_OFFSET_BOUND)
+        # The next thing this frame meets is an 8-bit quantisation, so a correction under
+        # 1/255 would be erased on the way there. Claiming it would be worse than silence.
+        if float((gain - 1.0).abs().max()) < 1e-3 and float(off.abs().max()) < 1.0 / 255.0:
+            return None, None
+        return gain, off
+
+    def note(self, gain, off):
+        """Record what was applied, and say it in one clause."""
+        self.applied.append((gain.clone(), off.clone()))
+        return (f"gain {'/'.join(f'{float(v):.3f}' for v in gain)} "
+                f"level {'/'.join(f'{float(v):+.4f}' for v in off)}")
+
+def _keyframe_latent(vae, hand_img):
+    """The keyframe latent for this shot: an ENCODE of the previous shot's last frame.
+
+    This was briefly an optimisation -- pass the previous shot's own latent straight
+    through and skip a VAE round trip per boundary. It was wrong, and it degraded
+    every shot after the first.
+
+    A keyframe is ONE pixel frame, and H3's grid puts that at 5f -> TWO latent
+    frames. Slicing [:, :, -1:] off a finished shot hands over one. Worse, the video
+    VAE is causal: the last latent of a 72-frame sequence encodes its temporal
+    context, not a standalone opening frame, so even at the right count it does not
+    mean what a keyframe means. The spatial-size guard could not see either problem.
+
+    The round trip is real but it is one lossy step on a correctly formed anchor,
+    which beats a cheap malformed one."""
+    return vae.encode(hand_img)
+
+
+def _build_ref_images(vae, images, gen_w, gen_h, mode="match"):
+    """(tokenizer items, DiT blocks) for a list of reference IMAGE tensors.
+
+    The tokenizer labels each one `<Picture N>:` itself, in the order given here --
+    so the roster the prompt refers to is decided by input order, not by anything
+    written in the prompt."""
+    items, blocks = [], []
+    for img in images:
+        if img is None:
+            continue
+        h, w = int(img.shape[1]), int(img.shape[2])
+        tw, th = ref_image_canvas(w, h, gen_w, gen_h, mode)
+        resized = _resize(img[:1], tw, th, "disabled")
+        items.append({"type": "image", "data": resized})
+        blocks.append({"kind": "image", "latent_h": th // 16, "latent_w": tw // 16,
+                       "latent": vae.encode(resized)})
+    return items, blocks
+
+
+def build_conditioning(clip, vae, audio_vae, prompt, width, height, length,
+                       handoff=None, refs=None,
+                       ref_noise_aug=0.999, silent=False, ref_image_size="match",
+                       handoff_as_ref=False, speech_lead_seconds=0.0, speech_tail_frames=0):
+    """Encode prompt, identity references, keyframe, and audio constraints for a shot."""
+    latent, fc = _empty_av_latent(width, height, length, H3_FPS)
+    refs = [r for r in (refs or []) if r is not None]
+
+    hand_img = None
+    if handoff is not None:
+        hand_img = _resize(handoff[:1], width, height, "disabled")
+
+    # REFERENCES AND THE KEYFRAME RIDE TOGETHER. This is the arrangement the node
+    # had before I broke it, and the reason is in ComfyUI's own layout:
+    #
+    #   model_base.py:2183-2191  cond_video_latents = keyframe latents THEN ref latents
+    #   model.py PackedLayout    emits keyframe "cond" segments THEN ref "ref_img" ones
+    #
+    # The two orders agree, so both channels coexist. A shot takes its references AND
+    # a real keyframe: the keyframe ANCHORS the first frame, which is what continuity
+    # needs, while a reference only supplies identity. They are not alternatives.
+    #
+    # I had read "<Picture 1>" as MEANING the first frame on fl2va, and rearranged the
+    # roster around that. It does not. Which image is the first frame is decided by
+    # resolved_frame_index in minimax_keyframes, not by a label's number -- the labels
+    # are only how the images are shown to the VLM, and what they have to line up with
+    # is the <Picture N> tags in the prompt.
+    #
+    # So references come FIRST and keep slots 1..N, which is what a sheet line's
+    # `Name: <Picture 1>, ...` points at, and the handoff is appended AFTER them where
+    # it disturbs no numbering. It has to be in the list at all because
+    # tokenize_with_weights is either/or: passing minimax_ref_items makes it ignore
+    # `images` outright, so leaving the handoff out means the VLM is never shown where
+    # the shot left off and re-imagines the scenery -- same place, new room.
+    keyframe_ok = ref_noise_aug is None or float(ref_noise_aug) >= KEYFRAME_SAFE_AUG
+    # One aug covers every visual condition row, references AND the keyframe. Below
+    # KEYFRAME_SAFE_AUG the keyframe latent would be noised and labelled at the wrong
+    # timestep, so the handoff stops being an anchor and rides as an extra reference
+    # instead: weaker continuity, but nothing pretending to anchor while carrying noise.
+    # ...or because the caller asked for it. A shot that introduces somebody already
+    # in position wants the room this picture carries and NOT the first frame it
+    # would force, and that is a demotion the aug knows nothing about.
+    carry_as_ref = bool(hand_img is not None
+                        and (handoff_as_ref or (refs and not keyframe_ok)))
+
+    enc_refs = refs + ([hand_img] if carry_as_ref else [])
+    items, blocks = ([], [])
+    if enc_refs:
+        items, blocks = _build_ref_images(vae, enc_refs, width, height, ref_image_size)
+    if hand_img is not None and not carry_as_ref:
+        items = items + [{"type": "image", "data": hand_img}]
+
+    if items:
+        tokens = clip.tokenize(prompt, minimax_ref_items=items)
+    else:
+        tokens = clip.tokenize(prompt)
+    cond = clip.encode_from_tokens_scheduled(tokens)
+
+    vals = {}
+    if blocks:
+        vals["minimax_refs"] = blocks
+        # How CLEAN the references are shown. One aug covers every conditioning
+        # latent, keyframe included -- which is why softening references below
+        # KEYFRAME_SAFE_AUG would soften the anchor too.
+        if ref_noise_aug is not None:
+            vals["minimax_visual_cond_noise_aug"] = float(ref_noise_aug)
+
+    kfs = []
+    if hand_img is not None and not carry_as_ref:
+        kfs.append({"resolved_frame_index": 0,
+                    "latent": _keyframe_latent(vae, hand_img)})
+    # Audio keyframes are extra conditioning rows in H3's PackedLayout. Pin the
+    # generated target stream instead, so the joint model also sees a quiet mouth.
+    # A dialogue shot pins its opening (the lead) and, past the line's estimated end,
+    # its close (the tail); the span between is the model's.
+    if silent or float(speech_lead_seconds or 0.0) > 0.0 or int(speech_tail_frames or 0) > 0:
+        _SILENCE_STATUS["asked"] += 1
+        if audio_vae is None:
+            _SILENCE_STATUS["why"] = "no audio VAE is wired to the node"
+        else:
+            sil = _silent_audio_latent(audio_vae, fc, H3_FPS)
+            if sil is None:
+                _SILENCE_STATUS["why"] = ("the audio VAE would not encode a silent "
+                                          "second -- the wrong VAE is on the "
+                                          "audio_vae input")
+            else:
+                lead = None if silent else round(float(speech_lead_seconds) *
+                                                 AUDIO_LATENT_FPS)
+                tail = 0 if silent else int(speech_tail_frames or 0)
+                if _pin_audio_silence(latent, sil, lead, tail):
+                    _SILENCE_STATUS["applied"] += 1
+                else:
+                    _SILENCE_STATUS["why"] = "the silent latent did not match the shot"
+    if kfs:
+        vals["minimax_keyframes"] = kfs
+    if vals:
+        cond = node_helpers.conditioning_set_values(cond, vals)
+    return cond, latent, fc, carry_as_ref

dumas_h3_longvideos_engine.py

@@ -66,21 +66,8 @@ HARDWARE = (
     (r"cuffs?|cuffed", "cuffs", "wrists"),
     (r"tape", "tape", "wrists"),
 )
-# Material and colour survive because they decide what the thing looks like:
-# "steel collar" must not come back as "collar" two shots later.
-# WHAT THE AUTHOR CALLED IT. This decides how much of the wording survives into
-# the guard clauses, and the guard is what every shot after the first repeats --
-# so a word missing here is a word the model stops hearing.
-#
-# It was twenty-odd words, and "a mirrored steel collar" came back as "steel
-# collar" while "a brushed nickel collar" came back as "collar". A bare "collar"
-# repeated once a shot is a bare collar, and the prior for that is a black
-# leather one -- which is exactly what was reported.
-#
-# Hyphenated compounds pass whole ("mirror-finish", "chrome-plated"), so an
-# unusual finish survives without being listed. Bare participles are NOT
-# accepted: "-ed" is a verb far more often than a modifier, and capturing one
-# would put an action into the name of the thing.
+# Preserve visual modifiers in continuity text. Hyphenated compounds pass whole;
+# arbitrary participles do not, because they are more often verbs than modifiers.
 _ADJ = (r"(?:[A-Za-z]+-[A-Za-z]+|"
         # materials
         r"steel|stainless|iron|metal|metallic|nickel|chrome|chromed|brass|"
@@ -131,8 +118,26 @@ PART_VARIES = frozenset({"chain", "rope", "straps", "tape"})
 # outlives the beat that caused it, and the clause that says so has to be
 # writable from any later shot.
 REGION_OF = (
+    # UNDERWEAR IS IN HERE TOO, on both halves of the body. The torso row has
+    # listed a bra since the day it was written -- that is the report it exists
+    # for, "a bra coming back on somebody topless" -- and the leg row never got
+    # its counterpart, so region_of("thong") answered "". A garment that cannot
+    # be placed latches no bare region, so underwear coming off said nothing
+    # about the hips in that shot or in any shot after it, and an unspecified
+    # region is filled by the model's own prior -- which for a hip is underwear.
+    #
+    # Worse, a beat saying somebody is NAKED looks each worn garment's region up
+    # to take it off the body, so the one garment it could not place stayed
+    # "worn" in the state while the text said she was nude. Reported as a thong
+    # restored a beat after she undressed to get in the shower.
+    #
+    # No hardware. A chastity belt is in the layering vocabulary, but it is a
+    # restraint: it is latched and held by its own mechanism, and a bare region
+    # read off it would argue with that.
     (r"shorts|trousers|jeans|slacks|chinos|skirt|kilt|leggings|joggers|tights|"
-     r"pantyhose|jeggings|culottes|tracksuit\s+bottoms", "legs",
+     r"pantyhose|jeggings|culottes|tracksuit\s+bottoms|"
+     r"panties|knickers|thong|g-?string|briefs|boxers|underwear|undies|"
+     r"jockstrap|loincloth", "legs",
      "The legs are bare from the hip down"),
     (r"socks|stockings|hold-?ups|boots|shoes|trainers|sneakers|sandals|heels",
      "feet", "The feet and ankles are bare"),
@@ -214,16 +219,34 @@ RELEASE_VERB = (
 # the engine knew a cell and a warehouse, the sampler did not, so a scene set in
 # either was a room to one reader and nowhere to the other. Same fault the
 # garment lists had, waiting to be reported.
+# MULTI-WORD ROOMS COME BEFORE BARE "room". _MOD is non-greedy, so it tries no
+# modifier first and the longest place wins -- but only if the long form is here to
+# win with. Without "locker\s+room", "heads to the locker room" read its destination
+# as "room": _MOD swallowed "locker" and the capture took what was left, so the film
+# was reported as entering "room" and the shot was told to arrive in one.
+#
+# A gym, a locker room and a court were in no list at all, which is worse than vague:
+# with the ORIGIN unknown, travel_anchor emitted nothing, so a beat walking out of a
+# gym was never told to walk and the set simply changed under the characters.
 PLACES = (r"hallway|hall|corridor|passage|landing|stairwell|staircase|stairs|"
           r"steps|bedroom|bathroom|washroom|kitchen|living\s+room|lounge|"
-          r"dining\s+room|study|office|garage|basement|cellar|attic|loft|porch|"
+          r"dining\s+room|locker\s+rooms?|changing\s+rooms?|dressing\s+rooms?|"
+          r"waiting\s+rooms?|utility\s+rooms?|gymnasium|gym|classroom|library|"
+          r"cafeteria|canteen|reception|laundry|pantry|sauna|balcony|terrace|"
+          r"rooftop|elevator|court|pool|showers|shower|store|shop|studio|"
+          r"study|office|garage|basement|cellar|attic|loft|porch|"
           r"veranda|garden|yard|driveway|street|alley|car\s?park|lobby|foyer|"
           r"doorway|cell|warehouse|barn|shed|van|truck|room")
 # Place words that are also ordinary verbs or everyday nouns. A reader with a
 # preposition in front of it ("in the study") can tell which sense is meant; the
 # free-text one cannot, and "she steps out", "they study the map" and "he lands
 # badly" are all commoner than the rooms they collide with.
-PLACE_ALSO_A_VERB = {"steps", "landing", "study", "lounge", "garage", "porch"}
+# "bar" and "lift" are deliberately NOT places in this file at all: bars are
+# restraint hardware here ("chained to the bars") and lifting is what happens to a
+# garment or a body, so behind a preposition they would both read as journeys.
+PLACE_ALSO_A_VERB = {"steps", "landing", "study", "lounge", "garage", "porch",
+                     "court", "pool", "shower", "showers", "store", "shop",
+                     "studio", "reception"}
 # A room is usually described, not just named -- "the tiled bathroom", "the long
 # hallway". Up to three adjectives, non-greedy so the NEAREST room still wins.
 _ROOM_MOD = (r"(?:(?!(?:of|the|an?|and|or|to|in|into|from|with|on|at|by|for|her|"
@@ -272,15 +295,21 @@ _GAP = r"(?:\s+\S+){0,4}?\s+"
 TAKES_OFF = (r"(?:takes?|took|taking|pulls?|pulled|peels?|peeled|strips?|"
              r"stripped|shrugs?|slips?|slipped|steps?|gets?|got|kicks?|"
              r"kicked)" + _GAP + r"(?:off|out\s+of)\b"
-             r"|\b(?:removes?|removed|removing|discards?|discarded|"
-             r"undresses|undressed|unbuttons?|unzips?|unzipped)")
+             r"|\b(?:removes?|removed|removing|discards?|discarded|sheds?|shedding|"
+             r"undresses|undressed)")
 PUTS_ON = (r"(?:puts?|putting|pulls?|pulled|slips?|slipped|tugs?|tugged|"
            r"steps?|stepped|climbs?|climbed|gets?|got|wriggles?)" + _GAP +
            r"(?:on|into|back\s+on)\b"
            r"|\b(?:dresses?\s+in|dressed\s+in|buttons?|zips?\s+up|fastens?)")
+# UNZIPPING A JACKET LEAVES IT ON. These were removals, so "Owen unzips his jacket"
+# took the jacket out of every later shot and called his chest bare -- and "rolls up
+# his sleeves" a beat later rolled the sleeves of a shirt that had gone with it. They
+# open a garment; a beat that also takes it off says so ("and takes it off").
 DISPLACES = (r"(?:pulls?|pulled|pushes?|pushed|tugs?|tugged|hikes?|hiked|"
              r"rolls?|rolled|lifts?|lifted|yanks?|yanked|shoves?|shoved)"
-             + _GAP + r"(?:aside|up|down|open)\b")
+             + _GAP + r"(?:aside|up|down|open)\b"
+             r"|\b(?:unzips?|unzipped|unbuttons?|unbuttoned|unfastens?|unfastened|"
+             r"undoes|undid)\b")
 
 # POSTURES and the _POSTURE list built from it used to live here. Nothing read
 # _POSTURE -- it was a second, dead copy of the posture vocabulary, and it had
@@ -440,6 +469,17 @@ def _outside_speech(text):
     return _SPOKEN_SPAN.sub(" ", text or "")
 
 
+def staged_text(text):
+    """What a beat STAGES: not what anybody says, and not what the narration asks.
+
+    A narrated question is the same kind of thing as a line of speech. "Maya waits by
+    the door. Will he come?" asks whether he will, which is to say he is not there --
+    and the "he" read as him being present, so Will was described into the shot of
+    her waiting for him. Removed for deciding who is in the shot, as speech is."""
+    staged = _outside_speech(text)
+    return " ".join(s for s in re.split(r"(?<=[.!?])\s+", staged) if not s.rstrip().endswith("?"))
+
+
 # THREE modifiers, not two: "mirrored stainless steel collar" is three words and
 # a noun, and the third was the first to be dropped.
 _HW_ONE = _rx(r"\b(" + _ADJ + r"(?:\s+" + _ADJ + r"){0,2}\s+)?("
@@ -478,6 +518,14 @@ _GARMENT_ONE = _rx(r"\b(" + _ADJ + r"(?:\s+" + _ADJ + r"){0,2}\s+)?("
 _TAKES_OFF = _rx(r"\b" + TAKES_OFF + r"\b")
 _PUTS_ON = _rx(r"\b" + PUTS_ON + r"\b")
 _DISPLACES = _rx(r"\b" + DISPLACES + r"\b")
+# ...and the sentence can still finish the job after the garment is named. "Kate
+# unzips the denim skirt and steps out of it": the unzip opens it, the rest of the
+# sentence takes it off, and the removal verb comes after the item where the reader
+# above does not look.
+_OPENS_GARMENT = _rx(r"\b(?:unzips?|unzipped|unbuttons?|unbuttoned|unfastens?|unfastened|"
+                     r"undoes|undid|unhooks?|unhooked|unclasps?|unclasped)\b")
+_COMPLETES_OFF = _rx(r"\b(?:off|out\s+of|away|lets?\s+(?:it|them)\s+(?:fall|drop|slide)|"
+                     r"drops?\s+(?:it|them)|falls?\s+(?:to|down|away|off))\b")
 _MOVES = _rx(r"\b(?:walks?|walked|walking|goes|go|went|going|runs?|ran|running|"
              r"steps?|stepped|stepping|moves?|moved|moving|enters?|entered|"
              r"leaves?|left|leaving|crosses|crossed|crossing|climbs?|climbed|"
@@ -595,11 +643,6 @@ def nudity_in(text):
     return out
 
 
-def bare_sentence(region):
-    """How to say a region is bare, or "" for one with no wording."""
-    return next((s for _rx, r, s in _REGION_RX if r == region), "")
-
-
 def _bare_on(p, regions):
     for r in ([regions] if isinstance(regions, str) else regions):
         if r and r not in p.bare:
@@ -653,11 +696,6 @@ def _nearest_part(parts, at, ats):
     return ""
 
 
-def hardware_in(text):
-    """Every piece of hardware named, as (canonical, part, as-written)."""
-    return [(c, p, w) for c, p, w, _at in hardware_spans(text)]
-
-
 def position_spans(text):
     """Every limb position named, as (name, at)."""
     out = []
@@ -705,9 +743,15 @@ def place_in(text):
 
     Behind a preposition, so a room has to be somewhere somebody IS. "Ana looks
     at the door" names no room -- and a door is not on the list in any case."""
-    m = _PLACE_IN.search(text or "")
+    text = text or ""
+    m = _PLACE_IN.search(text)
     if not m:
         return ""
+    clause = re.split(r"[.;!?]", text[:m.end()])[-1]
+    present = re.search(r"\b(?:is|are|was|were|stands?|sits?|waits?|lies?|remains?)\b",
+                        clause, re.I)
+    if not _MOVES.search(clause) and not present:
+        return ""
     got = re.sub(r"\s+", " ", m.group(1).lower()).strip()
     # A BARE "room" NAMES NOWHERE. "Ana walks into the room" says she goes
     # inside, not which room -- and taking it as a place produced "The shot is
@@ -730,6 +774,18 @@ def garments_in(text):
     return out
 
 
+_CLAUSE_BOUNDARY = re.compile(r"[,;]|\b(?:and|while)\b", re.I)
+
+
+def _clause_at(text, at, boundaries=None):
+    """Return the clause containing character offset `at` and its start offset."""
+    if boundaries is None:
+        boundaries = list(_CLAUSE_BOUNDARY.finditer(text or ""))
+    lo = max((m.end() for m in boundaries if m.end() <= at), default=0)
+    hi = min((m.start() for m in boundaries if m.start() > at), default=len(text))
+    return text[lo:hi], lo
+
+
 
 # Hardware, not clothing. Taking clothes off does not unlock anything, so these
 # are kept out of the garment answer -- the standing rule is that hardware is
@@ -740,6 +796,32 @@ _NOT_CLOTHING = re.compile(
     r"straitjacket|spreader|hogtie|clamps?|clips?)$", re.I)
 _PHRASE_ONE = _rx(r"\b(?:" + GARMENT_PHRASES + r")\b")
 _WORD_ONE = _rx(r"^(?:" + GARMENT_WORDS + r")s?$")
+# The same list with NO optional plural, which is what says whether a trailing "s"
+# belongs to the word or was added to it. The vocabulary is clean on this: garments that
+# are inherently plural are listed only in the plural (boots, jeans, shorts, panties,
+# tights, leggings, socks, knickers, trousers, gloves) and the rest only in the singular
+# (skirt, vest, top), so "stem is itself a garment" is an exact test and not a guess.
+_WORD_EXACT = _rx(r"^(?:" + GARMENT_WORDS + r")$")
+
+
+def singular_garment(word):
+    """A garment word as the VOCABULARY spells it, so two readers cannot disagree.
+
+    Reported: several women take their skirts off and the skirts are back in the next
+    beat. "their skirts" yields the token "skirts" while the sheet says "a denim
+    skirt", and every reader downstream looks the token up in the sheet -- the scrub
+    by pattern, infer_removals by entry head -- so a plural garment matched nothing
+    and the removal silently did nothing at all. One woman undressing wrote "her
+    skirt" and worked; the moment the subject went plural so did the garment.
+
+    A trailing "s" comes off only when the stem is ITSELF a garment word, which is an
+    exact test here and not a guess: the vocabulary lists inherently plural garments
+    only in the plural (boots, jeans, shorts, panties, tights, leggings, socks,
+    knickers, trousers, gloves) and the rest only in the singular."""
+    low = str(word or "").lower().strip("-")
+    if low.endswith("s") and _WORD_EXACT.match(low[:-1]):
+        return low[:-1]
+    return low
 
 
 def garment_words(text):
@@ -761,12 +843,60 @@ def garment_words(text):
     text = _PHRASE_ONE.sub(" ", text)
     for word in re.findall(r"\b[\w-]{3,}\b", text):
         low = word.lower().strip("-")
-        if low in out or _NOT_CLOTHING.match(low):
+        if _NOT_CLOTHING.match(low):
             continue
-        if _WORD_ONE.match(low):
+        if not _WORD_ONE.match(low):
+            continue
+        # The token the sheet wrote, not the one the beat happened to inflect. See
+        # singular_garment -- shared with infer_removals so the two cannot disagree.
+        low = singular_garment(low)
+        if low not in out:
             out.append(low)
     return out
 
+# A POSTURE DENIED IS NOT A POSTURE TAKEN.
+#
+# Reported: a woman chained by the ankles and forced into a squat stood up anyway.
+# The chain clauses were all correct -- the metal "already drawn to its full length,
+# so the position it fixes is the position that keeps" was on every shot after the
+# squat. What sat beside it was the posture latch saying she was STANDING, because
+# "She cannot stand." matched `stand` and nothing looked at the `cannot`. The latch
+# then carried that forward, so every later shot asserted, flatly and positively,
+# the one thing the chains were there to prevent. At cfg 1 a positive statement wins.
+#
+# The same shape as _in_a_request, which already suppresses a posture that is ASKED
+# for rather than taken. Attempts are here too: "tries to stand", "struggles to get
+# up", "strains to rise" are all bodies that have NOT got there, and reading them as
+# arrival is the same error in a friendlier disguise.
+#
+# A SHORT WINDOW on purpose -- the five words before the verb. The cue always sits
+# immediately in front of it ("cannot stand", "no longer able to stand"), and a wider
+# reach would let a `cannot` from a different clause silence a real posture.
+_POSTURE_DENIED = _rx(
+    r"\b(?:cannot|can\s*not|can['\u2019]?t|could\s*not|could\s*n['\u2019]?t|"
+    r"unable|never|not\s+able|no\s+longer\s+able|"
+    r"does\s*n['\u2019]?t|does\s+not|do\s*n['\u2019]?t|did\s*n['\u2019]?t|did\s+not|"
+    r"will\s+not|wo\s*n['\u2019]?t|fail(?:s|ed|ing)?|"
+    r"tr(?:y|ies|ied|ying)|attempt(?:s|ed|ing)?|struggl(?:e|es|ed|ing)|"
+    r"strain(?:s|ed|ing)?|fight(?:s|ing)?|want(?:s|ed)?|need(?:s|ed)?|"
+    r"told|ordered|asked|begs?|begged)\b")
+
+
+def denied_posture(text, at):
+    """Is the posture verb at `at` negated, or only attempted, by what precedes it?"""
+    before = str(text or "")[:max(0, int(at))]
+    # A cue belongs to its OWN clause. Without stopping at the boundary, "McKenna
+    # cannot kneel, so she sits" reached back past the comma and silenced the sitting
+    # -- suppressing a posture the beat plainly states, which is the same class of
+    # error in the other direction.
+    cut = 0
+    for _m in re.finditer(r"[,;:.!?]|\b(?:so|and|but|then|yet|while|as|before|after)\b",
+                          before, re.I):
+        cut = _m.end()
+    window = " ".join(re.findall(r"[\w'\u2019]+", before[cut:])[-5:])
+    return bool(_POSTURE_DENIED.search(window))
+
+
 def posture_in(text):
     """The posture this beat puts a body in. '' when it does not.
 
@@ -775,7 +905,7 @@ def posture_in(text):
     for that, but off the same vocabulary."""
     t = text or ""
     hits = sorted((m.start(), name) for name, rx in _POSTURE_OF
-                  for m in [rx.search(t)] if m)
+                  for m in rx.finditer(t) if not denied_posture(t, m.start()))
     return hits[0][1] if hits else ""
 
 
@@ -820,7 +950,7 @@ _TRAILING_VERB = (r"take[sn]?|took|taking|pull(?:s|ed|ing)?|peel(?:s|ed|ing)?|"
                   r"toss(?:es|ed)?|throw[s]?|threw|kick(?:s|ed|ing)?|"
                   r"slide[s]?|slid|wriggle[sd]?|wiggle[sd]?")
 # ...and verbs that are a removal on their own, needing no particle.
-_UNDO_VERB = (r"remove[sd]?|removing|undress(?:es|ed)?|unzip(?:s|ped)?|"
+_UNDO_VERB = (r"remove[sd]?|removing|undress(?:es|ed)?|shed(?:s|ding)?|unzip(?:s|ped)?|"
               r"unbutton(?:s|ed)?|unhook(?:s|ed)?|unclasp(?:s|ed)?|unfasten(?:s|ed)?|"
               # Hardware comes off by being UNDONE, and these were missing: a beat
               # saying "unlocks the belt" left it described as worn for the rest of
@@ -830,11 +960,26 @@ _UNDO_VERB = (r"remove[sd]?|removing|undress(?:es|ed)?|unzip(?:s|ped)?|"
               r"undo(?:es)?|undid")
 
 
+# WHERE A GARMENT ENDS UP ONCE IT IS OFF. A thing on the floor is not on a body,
+# and this is how a beat says so -- by DESTINATION, not by verb. Two readers need
+# the same list, for opposite reasons: the removal reader to call it a removal,
+# the restore reader to stop calling it one. "Lets the skirt fall" drops a lifted
+# skirt back over her legs; "lets the thong fall to the floor" is the thong coming
+# off, and the only difference between those two sentences is this list.
+FLOOR = (r"floor|ground|tiles?|tiling|lino|mat|bath\s*mat|rug|carpet|deck|boards|"
+         r"concrete|grass|sand|bed|sofa|couch|chair|seat|stool|bench|basket|"
+         r"hamper|laundry|pile|heap")
+TO_THE_FLOOR = (r"(?:to|on|onto|into|in)\s+(?:the|a|an|her|his|their)?\s*"
+                r"(?:" + FLOOR + r")\b")
+_LANDS_OFF = _rx(r"\s*(?:fall(?:s|ing)?|drop(?:s|ping)?|land(?:s|ing)?)?\s*"
+                 + TO_THE_FLOOR)
+
 _DISPLACE_WAY = (r"back\s+up|back\s+down|down|up|aside|open|back|"
                  r"off\s+(?:one|her|his|their)\s+shoulders?")
 _DISPLACE = re.compile(
     r"\b(?:" + _STRIP_VERB + r"|push(?:es|ed|ing)?|shove[sd]?|roll(?:s|ed|ing)?|"
-    r"hitch(?:es|ed)?|hike[sd]?|open(?:s|ed)?|undo(?:es)?|unzip(?:s|ped)?|"
+    r"hitch(?:es|ed)?|hike[sd]?|open(?:s|ed)?|undo(?:es)?|undid|unzip(?:s|ped)?|"
+    r"unbutton(?:s|ed)?|unfasten(?:s|ed)?|unhook(?:s|ed)?|unclasp(?:s|ed)?|"
     # LIFTING A SKIRT IS DISPLACING IT, and none of these were here. Asked
     # for directly: "when the skirt has been lifted up to show the chastity
     # belt, that's when it should be shown". Lifting was not read as moving
@@ -874,14 +1019,23 @@ def scene_name_for(head, scene):
             # a reference is pinning, so it is the worst one to describe loosely.
             item = re.sub(r"<\s*picture\s+\d+\s*>", " ", item, flags=re.I)
             item = re.sub(r"\s+", " ", item).strip()
-            if not item or item.split()[-1].lower() != head:
+            # ONE ENTRY CAN HOLD SEVERAL GARMENTS, and the name is the garment's own
+            # part of it. "navy jacket over a white shirt" ends in "shirt", so the
+            # whole entry came back as the shirt's name and "the navy jacket over a
+            # white shirt open" was said about a shirt being unbuttoned. Split on the
+            # words that join garments, and cut a "with ..." tail, which describes a
+            # garment rather than naming it.
+            for part in re.split(r"\s+(?:over|under|beneath|underneath|on\s+top\s+of|and)\s+",
+                                 item, flags=re.I):
+                part = re.split(r"\s+with\s+", part, flags=re.I)[0].strip()
+                if not part or part.split()[-1].lower() != head:
                     continue
                 # Drop a leading article or possessive; they are not description.
-            item = re.sub(r"^(?:a|an|the|her|his|their|its)\s+", "", item, flags=re.I)
-            # The longest entry wins: a sheet that names it twice described it most
-            # fully once, and the fuller name is the one worth carrying.
-            if len(item) > len(best):
-                best = item
+                part = re.sub(r"^(?:a|an|the|her|his|their|its)\s+", "", part, flags=re.I)
+                # The longest entry wins: a sheet that names it twice described it
+                # most fully once, and the fuller name is the one worth carrying.
+                if len(part) > len(best):
+                    best = part
     # The author's OWN capitalisation. Lowercasing turned "PVC" into "pvc" and
     # "Shiny white crop top" into all-lowercase -- a different token sequence than
     # was written, for a brand or material name that is capitalised for a reason.
@@ -910,6 +1064,13 @@ def displaced_garments(beat, scene):
         if not way and re.match(r"\s*(?:lift|rais|hoist|gather|bunch)", m.group(0),
                                 re.I):
             way = "up"
+        # ...and UNDOING a garment opens it. "Owen unzips his jacket" was a removal,
+        # then (once it was not) nothing at all -- the jacket went back to being
+        # described closed on the next shot, which is a jacket zipping itself up
+        # across a cut.
+        if not way and re.match(r"\s*(?:unzip|unbutton|unfasten|unhook|unclasp|undo|undid)",
+                                m.group(0), re.I):
+            way = "open"
         if not way or not thing or thing in seen:
             continue
         # The garment has to be one the scene already dresses them in, and the head
@@ -987,9 +1148,26 @@ def restored_garments(beat, scene):
         return []
     out, low = [], scene.lower()
     for m in _PUT_BACK_NAMED.finditer(beat):
+        # ...UNLESS IT LANDS ON THE FLOOR. These verbs are the restore vocabulary
+        # because that is what people write for a lifted skirt -- let fall, drop,
+        # lower, let go of -- and the identical words take a garment OFF when the
+        # sentence says where it lands. "Lets the thong fall to the floor" was read
+        # as putting the thong back on: the author's removal, enacted backwards, and
+        # from there the sheet described it as worn for the rest of the film.
+        # Reported as a thong restored after she undressed. See FLOOR.
+        if _LANDS_OFF.match(beat[m.end():]):
+            continue
         thing = re.sub(r"\s+", " ", (m.group(1) or "")).strip().lower()
         if not thing:
             continue
+        # ...AND THE CAPTURE CANNOT RUN THROUGH A PREPOSITION. The group takes
+        # spaces so a sheet's "long grey skirt" comes back whole, and on "drops the
+        # thong on the floor" it swallowed "thong on the floor" instead -- head
+        # "floor" -- so the restore was keyed to the ROOM. The scene named a wet
+        # floor, scene_name_for handed back "tiled bathroom with a wet floor", and
+        # the beat was recorded as putting the bathroom back on.
+        if re.search(r"\b(?:on|onto|to|into|in|at|over|under|from|with|and)\b", thing):
+            continue
         head = thing.split()[-1]
         if len(head) < 3 or head not in low:
             continue
@@ -1197,6 +1375,10 @@ def _alias_at(word, staged):
     return fallback
 
 
+_AUX_FOLLOWER = re.compile(r"\s+(?:I|you|he|she|we|they|it|there|this|that|anyone|someone|"
+                           r"everyone|anybody|somebody)\b")
+
+
 def names_in(beat, cast):
     """Names this beat STAGES, in the order the sentence puts them.
 
@@ -1208,14 +1390,23 @@ def names_in(beat, cast):
     Speech-stripped, for the same reason it is everywhere else -- "McKenna, where
     are you?" is how absence gets written, and reading it as presence put a whole
     sheet entry into a shot the person is not in."""
-    staged = _outside_speech(beat or "")
+    staged = staged_text(beat or "")
     names = [str(n) for n in (cast or []) if n]
     hits, found = [], set()
     for n in names:
-        m = re.search(r"\b" + re.escape(n) + r"\b", staged)
-        if m:
+        # A NAME THAT IS ALSO A WORD. "Will he come?" and "May I come in?" open a
+        # sentence with the name followed by who the question is about, and read as
+        # the person they staged Will and May -- a second character in a shot about
+        # somebody waiting for them. A name followed straight away by a subject
+        # pronoun at the start of a sentence is the verb; the next use can still be
+        # the person ("Will opens the gate").
+        for m in re.finditer(r"\b" + re.escape(n) + r"\b", staged):
+            _opens = re.search(r"(?:^|[.!?]\s*[\"'\u201c]?)\s*$", staged[:m.start()])
+            if _opens and _AUX_FOLLOWER.match(staged[m.end():]):
+                continue
             hits.append((m.start(), n))
             found.add(n)
+            break
     # A SHEET NAME IS OFTEN LONGER THAN WHAT THE BEATS CALL HER. "Mistress Vale"
     # on the sheet and "the Mistress" in every beat matched nothing, so her line
     # was in no shot at all and the model invented her from scratch each time.
@@ -1415,20 +1606,35 @@ class SceneState:
         subject = who[0] if who else next(iter(list(self.people) or list(cast)
                                                or [""]))
 
-        hw = hardware_in(beat)
-        applying = bool(hw) and bool(_APPLY.search(beat)) and not _RELEASE.search(beat)
+        spans = hardware_spans(beat)
+        garments = list(_GARMENT_ONE.finditer(beat))
+        boundaries = list(_CLAUSE_BOUNDARY.finditer(beat)) if spans or garments else []
+        applying = bool(spans) and bool(_APPLY.search(beat))
         releasing = bool(_RELEASE.search(beat))
 
-        if applying:
-            wearer = _wearer(beat, who, subject)
-            p = self.person(wearer)
+        if applying or releasing:
             # MODIFIERS BIND TO THE NEAREST ITEM. "handcuffs her wrists behind
             # her back and locks a steel collar around her neck, chained to the
             # wall" carries two modifiers and two items; giving both modifiers
             # to both items produced handcuffs chained to a wall they were never
             # near, and a collar held behind a back.
-            spans = hardware_spans(beat)
             for canon, part, written, at in spans:
+                clause, lo = _clause_at(beat, at, boundaries)
+                item_at = at - lo
+                apply_at = max((m.start() for m in _APPLY.finditer(clause)
+                                if m.start() <= item_at), default=-1)
+                release_at = max((m.start() for m in _RELEASE.finditer(clause)
+                                  if m.start() <= item_at), default=-1)
+                if apply_at < 0 and release_at < 0:
+                    continue
+                local_who = names_in(clause, cast)
+                wearer = _wearer(clause, local_who or who, subject)
+                p = self.person(wearer)
+                if release_at >= 0:
+                    keys = [k for k in list(p.hardware) if k[0] == canon]
+                    for key in keys:
+                        changed["released"].append((wearer, p.hardware.pop(key)))
+                    continue
                 # KEYED BY THE PAIR. A chain on the ankles and a chain on the
                 # wrists are two restraints; keyed by name alone the second
                 # overwrote the first and one of them was never drawn again.
@@ -1446,7 +1652,7 @@ class SceneState:
             # Its anchor needs no transferring either: with the tether gone from
             # the spans, the anchor binds to the nearest remaining item, which
             # is the one it was always describing.
-        elif releasing:
+        if releasing and not spans:
             # Whoever is actually wearing it. "The guard unlocks the handcuffs"
             # names only the agent, and taking the subject there tried to
             # release hardware from the man holding the key.
@@ -1457,12 +1663,7 @@ class SceneState:
             # Released by NAME, whatever part it is on: an unlocking beat says
             # "unlocks the chain", not which of two chains, and matching the
             # pair left one fastened forever.
-            _kinds = {c for c, _pt, _w in hw}
-            named = [k for k in list(p.hardware) if k[0] in _kinds]
-            if named:
-                for key in named:
-                    changed["released"].append((wearer, p.hardware.pop(key)))
-            elif re.search(r"\b(?:them|it|her|him|everything|all\s+of\s+it)\b",
+            if re.search(r"\b(?:them|it|her|him|everything|all\s+of\s+it)\b",
                          beat, re.I):
                 # "the guard releases her" names no item, so all of it comes off.
                 while p.hardware:
@@ -1472,12 +1673,26 @@ class SceneState:
         # to be named -- a bare "she undresses" says nothing about which garment,
         # and guessing is how a garment came off a beat before the beat that
         # took it off.
-        wearer_g = _wearer(beat, who, subject) if len(who) > 1 else subject
-        if wearer_g:
-            p = self.person(wearer_g)
-            for g in garments_in(beat):
+        if subject:
+            for m in garments:
+                g = f"{(m.group(1) or '').strip()} {m.group(2)}".strip().lower()
                 key = _garment_key(g)
-                if _TAKES_OFF.search(beat):
+                clause, lo = _clause_at(beat, m.start(), boundaries)
+                item_at = m.start() - lo
+                actions = [(x.start(), "off") for x in _TAKES_OFF.finditer(clause)
+                           if x.start() <= item_at]
+                actions += [(x.start(), "on") for x in _PUTS_ON.finditer(clause)
+                            if x.start() <= item_at]
+                actions += [(x.start(), "aside") for x in _DISPLACES.finditer(clause)
+                            if x.start() <= item_at]
+                action = max(actions, default=(-1, ""))[1]
+                if action == "aside" and _OPENS_GARMENT.search(clause[:item_at]):
+                    if _COMPLETES_OFF.search(re.split(r"[.;!?]", beat[m.end():])[0]):
+                        action = "off"
+                local_who = names_in(clause, cast)
+                wearer_g = _wearer(clause, local_who, subject)
+                p = self.person(wearer_g)
+                if action == "off":
                     if key not in [_garment_key(x) for x in p.removed]:
                         p.removed.append(g)
                         changed["removed"].append((wearer_g, g))
@@ -1485,7 +1700,7 @@ class SceneState:
                     p.displaced = [x for x in p.displaced
                                    if _garment_key(x) != key]
                     _bare_on(p, region_of(g))
-                elif _PUTS_ON.search(beat):
+                elif action == "on":
                     if key not in [_garment_key(x) for x in p.worn]:
                         p.worn.append(g)
                         changed["worn"].append((wearer_g, g))
@@ -1496,7 +1711,7 @@ class SceneState:
                     # dresses is told for the rest of the film that the region is
                     # bare, over the garment she just put on.
                     _bare_off(p, region_of(g))
-                elif _DISPLACES.search(beat):
+                elif action == "aside":
                     if key not in [_garment_key(x) for x in p.displaced]:
                         p.displaced.append(g)
                         changed["displaced"].append((wearer_g, g))
@@ -1506,7 +1721,11 @@ class SceneState:
         # ever said what was on the chest.
         _nude = nudity_in(beat)
         if _nude:
-            for n in (who or ([subject] if subject else [])):
+            nude_at = min((m.start() for rx, _regions in _NUDITY_RX
+                           for m in rx.finditer(beat)), default=len(beat))
+            located = [(abs(beat.find(n) - nude_at), n) for n in who if beat.find(n) >= 0]
+            owners = [min(located)[1]] if located else ([subject] if subject else [])
+            for n in owners:
                 q = self.person(n)
                 _bare_on(q, _nude)
                 # ...and it takes the garments OFF. Saying somebody is topless

dumas_h3_longvideos_runtime.py

@@ -0,0 +1,739 @@
+# H3-LongVideos -- https://github.com/Smite79/MiniMax-H3-LongVideos
+# Copyright (c) 2026 Smite79. All rights reserved.
+# Redistribution, in whole or in part, requires written permission.
+# This notice may not be removed or altered. See LICENSE.
+"""Sampling, decoding, resizing, memory handling, and frame assembly."""
+
+import math
+
+import torch
+import nodes
+import comfy.utils
+import comfy.sample
+import comfy.samplers
+import comfy.nested_tensor
+import comfy.model_management as mm
+import latent_preview
+
+
+class FrameAccumulator:
+    """Build the final frame tensor once, retaining overflow only when necessary."""
+
+    def __init__(self, capacity, dtype, store_on_cpu):
+        self.capacity = int(capacity)
+        self.dtype = dtype
+        self.store_on_cpu = bool(store_on_cpu)
+        self.tensor = None
+        self.used = 0
+        self.overflow = []
+
+    def add(self, frames):
+        count = int(frames.shape[0])
+        if self.tensor is None and count:
+            device = torch.device("cpu") if self.store_on_cpu else frames.device
+            self.tensor = torch.empty(
+                (max(count, self.capacity),) + tuple(frames.shape[1:]),
+                dtype=self.dtype, device=device)
+        if (not self.overflow and self.tensor is not None
+                and self.used + count <= self.tensor.shape[0]):
+            self.tensor[self.used:self.used + count].copy_(frames)
+            self.used += count
+            return
+        self.overflow.append(frames.to("cpu", self.dtype, copy=True)
+                             if self.store_on_cpu else frames)
+
+    def release(self):
+        """Drop every tensor held, now, rather than whenever the collector gets to it.
+
+        On an interrupt the render unwinds through frames the collector tears down in
+        its own order, and a large video buffer freed after the models it was sized
+        against have already gone is a free the allocator cannot explain. Deliberately
+        does NOT empty the cache: that is another CUDA call, and if the context is
+        already in a sticky error state it is one more thing to abort inside."""
+        self.tensor = None
+        self.overflow = []
+        self.used = 0
+
+    def finish(self):
+        if not self.overflow:
+            if self.tensor is None:
+                return torch.cat(self.overflow, dim=0)
+            if self.used == self.tensor.shape[0]:
+                out = self.tensor
+            else:
+                # COMPACT, never a slice. A slice of a larger buffer keeps the WHOLE
+                # buffer's storage alive, which is the retention this class exists to
+                # prevent -- and test_the_chain_is_never_held_twice measures exactly
+                # that, demanding no unused bytes behind the returned tensor.
+                #
+                # There is slack because the capacity is now an upper bound: it can no
+                # longer assume trim_seam drops a frame at every seam, since a shot that
+                # opens on no keyframe keeps its first frame. Over-allocating by at most
+                # one frame per seam and compacting once is the bounded cost. The
+                # alternative -- an exact guess that can be too small -- drops into the
+                # overflow list, which with cleanup_between_shots off holds every shot's
+                # decoded frames live on the GPU until the end of the run.
+                out = torch.empty((self.used,) + tuple(self.tensor.shape[1:]),
+                                  dtype=self.dtype, device=self.tensor.device)
+                out.copy_(self.tensor[:self.used])
+            self.tensor = None
+            return out
+
+        extra = sum(int(piece.shape[0]) for piece in self.overflow)
+        reference = self.tensor if self.tensor is not None else self.overflow[0]
+        out = torch.empty((self.used + extra,) + tuple(reference.shape[1:]),
+                          dtype=self.dtype, device=reference.device)
+        if self.tensor is not None and self.used:
+            out[:self.used].copy_(self.tensor[:self.used])
+        at = self.used
+        while self.overflow:
+            piece = self.overflow.pop(0)
+            count = int(piece.shape[0])
+            out[at:at + count].copy_(piece)
+            at += count
+        self.tensor = None
+        return out
+
+
+H3_FPS = 24                    # H3 renders 24 fps, always
+
+
+AUDIO_LATENT_FPS = 40          # audio latent frames per second
+
+
+AUTO_TILE_T = 8                # temporal chunk for a tiled decode
+
+
+MAX_FRAMES = 362               # H3's own ceiling (~15s)
+
+
+CANVAS_MULTIPLE = 32
+
+
+REF_IMAGE_SHORT_EDGE = 2048
+
+
+def align_frame_count(n):
+    """Up to the next valid H3 frame count. The grid is 17k+5."""
+    n = max(5, int(n))
+    while n % 17 != 5:
+        n += 1
+    return min(n, MAX_FRAMES)
+
+
+def video_latent_t(fc):
+    return 2 if fc <= 5 else ((fc - 5) // 17) * 5 + 2
+
+
+def temporal_shape(length, fps=H3_FPS):
+    """(frame count, video latent frames, audio latent frames) for a shot.
+
+    `fps` is accepted but deliberately IGNORED: the audio latent has to line up
+    with 24 fps video or the shot's sound is stretched against its picture."""
+    fc = align_frame_count(length)
+    return fc, video_latent_t(fc), round(fc / H3_FPS * AUDIO_LATENT_FPS)
+
+
+def ref_image_canvas(w, h, gen_w, gen_h, mode="match"):
+    """Pure: the (width, height) a reference image is encoded at.
+
+    'match' scales it (DOWN only, aspect kept) to the generation's pixel area, so a
+    reference costs about as much as one frame of the shot. 'max' goes to the
+    reference pipeline's 2048 short edge for the best identity fidelity, which on a
+    long chain is several times slower because the rows are re-attended every step
+    of every shot. Never upscales: a small reference stays small."""
+    w, h = max(1, int(w)), max(1, int(h))
+    if mode == "max":
+        scale = min(1.0, REF_IMAGE_SHORT_EDGE / min(w, h))
+    else:
+        scale = min(1.0, math.sqrt((int(gen_w) * int(gen_h)) / float(w * h)))
+    snap = lambda v: max(CANVAS_MULTIPLE, round(v * scale / CANVAS_MULTIPLE) * CANVAS_MULTIPLE)
+    return snap(w), snap(h)
+
+
+def _resize(image, width, height, crop):
+    s = image[..., :3].movedim(-1, 1)
+    s = comfy.utils.common_upscale(s, width, height, "lanczos", crop)
+    return s.movedim(1, -1)
+
+
+def _empty_av_latent(width, height, length, fps, batch_size=1):
+    fc, lt, at = temporal_shape(length, fps)
+    video = torch.zeros([batch_size, 24, lt, height // 16, width // 16], device=mm.intermediate_device())
+    audio = torch.zeros([batch_size, 32, 2, at], device=mm.intermediate_device())
+    return {"samples": comfy.nested_tensor.NestedTensor((video, audio))}, fc
+
+
+def _auto_tile_t(n_latent_frames, requested=None):
+    """Temporal tile for a tiled decode. An explicit value wins.
+
+    The decode_tile_frames widget is gone, so this is where the value comes from
+    now. It has to come from somewhere: ComfyUI's decode_tiled_3d defaults tile_t
+    to 999, i.e. SPATIAL tiles only, and expanding the whole clip's time axis at
+    once is the single largest allocation in a run. A "tiled" decode that keeps the
+    full temporal extent barely lowers the peak, so the OOM retry that switches
+    tiling on was, without this, retrying with almost the same footprint."""
+    if requested:
+        return int(requested)
+    n = int(n_latent_frames or 0)
+    return AUTO_TILE_T if n > AUTO_TILE_T else None
+
+
+def _decode_video(vae, out_latent, tiled, free_first=None, tile_t=None, tile_xy=None,
+                  keep=()):
+    """Decode the video latent.
+
+    `free_first` is the diffusion model: sampling is finished, and the video VAE
+    needs the room for THIS decode -- the free runs immediately before it, not to
+    make room for the next shot. On a card where the DiT is most of the VRAM, the
+    decode does not fit until it goes.
+
+    `keep` is what must NOT be evicted on the way. It was `keep_loaded=[]`, which
+    unloaded every resident model -- including the video VAE, which ComfyUI then
+    reloaded three lines later to run the decode. An evict-and-reload of the thing
+    about to be used, once per shot, on every card. Peak VRAM is identical either
+    way, since the VAE has to be resident to decode; the round trip was pure cost.
+
+    memory_required is ASKED FOR HONESTLY, which it was not. It was 1e30, and
+    free_memory computes `memory_to_free = memory_required - get_free_memory(device)`
+    (model_management.py:887), so 1e30 means "unload everything not in keep_loaded",
+    every shot, in full -- skipping partially_unload entirely.
+
+    What that evicts is the DiT, three lines before the next shot needs it again. On
+    a machine whose RAM is already full of finished frames there is nowhere for it to
+    go but disk, so the reload is a read from the drive, once per shot. Reported as
+    thrashing that slows the preload, and it is exactly that: the same weights being
+    read back at every boundary.
+
+    The VAE knows what its own decode costs -- ComfyUI sizes it with
+    memory_used_decode and uses that number everywhere else. Asked for that instead,
+    a card with headroom frees NOTHING and the DiT simply stays. A card without
+    headroom frees what it needs and no more, which is what partially_unload is for.
+    1e30 remains the fallback for a VAE that cannot estimate itself."""
+    latent = out_latent["samples"]
+    if latent.is_nested:
+        latent = latent.unbind()[0]
+    if free_first is not None:
+        try:
+            mm.free_memory(_decode_headroom(vae, latent), mm.get_torch_device(),
+                           keep_loaded=_resident(keep or (vae,)))
+        except Exception:
+            pass
+    # A VAE THAT ALREADY TILES DOES NOT NEED TO BE ASKED TO, AND ASKING COSTS 3x.
+    #
+    # MiniMaxH3VideoVAE.decode_tiled is, in full:
+    #
+    #     def decode_tiled(self, z, **kwargs):
+    #         return self.decode(z)
+    #
+    # Every tile_t/overlap_t/tile_x/tile_y this function computes is discarded, so
+    # the tiling the widget promises is not happening here -- the model tiles
+    # internally either way (256px spatial, 17-frame temporal), which is why
+    # comfy/sd.py sets handles_tiling on it.
+    #
+    # What the detour costs is the OUTPUT BUFFER. comfy's VAE.decode preallocates
+    # ONE result at vae_output_dtype and hands it to the model as output_buffer=,
+    # and MiniMaxH3VideoVAE.decode_temporal writes finalized chunks straight into
+    # it. Going through decode_tiled instead reaches _decode_tiled_owned, which
+    # calls the model with output_buffer=None -- so decode_temporal allocates its
+    # own at torch.float32 -- and then makes an fp16 `copy=True` of that. Two
+    # buffers, the larger of them at double width:
+    #
+    #     tiled : fp32 2.60GB + fp16 copy 1.30GB = 3.90GB per shot
+    #     decode: one preallocated fp16          = 1.30GB per shot
+    #
+    # at 362 frames of 1056x608. Every shot, on the node's own default.
+    #
+    # So: when the VAE owns its tiling AND can be written into, the un-tiled call IS
+    # the tiled one, minus the copies. Anything else keeps the old path -- this is a
+    # detour around a detour, not a claim that tiling is useless.
+    _owns_tiling = bool(getattr(vae, "handles_tiling", False) and getattr(
+        getattr(vae, "first_stage_model", None), "comfy_has_chunked_io", False))
+    if tiled and _owns_tiling:
+        imgs = vae.decode(latent)
+    elif tiled:
+        # Temporal + spatial tiling. Without tile_t the VAE expands the WHOLE latent
+        # clip at once, which on a 243-frame 1344x768 shot is the single largest
+        # allocation in the run -- and on an unpruned checkpoint that is already
+        # streaming, it is what tips the card over. Decoding in temporal chunks
+        # trades a little speed for a much lower peak; None keeps ComfyUI's defaults.
+        args = {}
+        tile_t = _auto_tile_t(latent.shape[2] if latent.ndim >= 5 else 0, tile_t)
+        if tile_t:
+            args["tile_t"] = int(tile_t)
+            args["overlap_t"] = max(1, int(tile_t) // 8)
+        if tile_xy:
+            args["tile_x"] = int(tile_xy)
+            args["tile_y"] = int(tile_xy)
+        try:
+            imgs = vae.decode_tiled(latent, **args) if args else vae.decode_tiled(latent)
+        except TypeError:
+            imgs = vae.decode_tiled(latent)      # older signature without tile_t
+    else:
+        imgs = vae.decode(latent)
+    if len(imgs.shape) == 5:
+        imgs = imgs.reshape(-1, imgs.shape[-3], imgs.shape[-2], imgs.shape[-1])
+    return imgs
+
+
+def _decode_audio(audio_vae, out_latent):
+    latent = out_latent["samples"]
+    if latent.is_nested:
+        latent = latent.unbind()[-1]
+    audio = audio_vae.decode(latent).movedim(-1, 1)
+    std = torch.std(audio, dim=[1, 2], keepdim=True) * 5.0
+    std[std < 1.0] = 1.0
+    audio = audio / std
+    sr = getattr(audio_vae, "audio_sample_rate_output", getattr(audio_vae, "audio_sample_rate", 44100))
+    return {"waveform": audio, "sample_rate": sr}
+
+
+def _is_oom(e):
+    return isinstance(e, torch.cuda.OutOfMemoryError) or "out of memory" in str(e).lower()
+
+
+def _deep_cleanup():
+    """Release cached VRAM between shots so a long chain does not accumulate and OOM.
+
+    It unloads NOTHING. soft_empty_cache(force) ignores `force` in current ComfyUI
+    (model_management.py:2050) -- the body only reaches empty_cache() and
+    ipc_collect() -- so this drops cached blocks, not models. The `True` is kept
+    only for older builds that read it; the older comment here claimed this took an
+    unload_all_models path, and it does not."""
+    try:
+        mm.soft_empty_cache(True)
+    except TypeError:
+        mm.soft_empty_cache()
+    try:
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+            torch.cuda.ipc_collect()
+    except Exception:
+        pass
+
+
+DECODE_HEADROOM = 1.25          # over ComfyUI's own estimate, for working allocations
+
+
+SAMPLE_HEADROOM = 1.35          # likewise for sampling, which is the longer stretch
+
+
+def _decode_headroom(vae, latent):
+    """VRAM this decode actually needs, by the VAE's own estimate. 1e30 if unknown.
+
+    ComfyUI sizes every VAE with memory_used_decode and uses that number itself, so
+    it is the honest figure to hand free_memory. The alternative -- and what was here
+    -- is 1e30, which means "unload everything" and evicts the DiT before every
+    decode, three lines before the next shot reloads it.
+
+    1e30 on failure rather than 0: a bad estimate that frees too little turns a slow
+    render into an OOM, and a wrong guess should fall back to the behaviour that has
+    been running, not to no freeing at all."""
+    try:
+        dtype = getattr(vae, "vae_dtype", None) or latent.dtype
+        need = float(vae.memory_used_decode(tuple(latent.shape), dtype))
+        if need > 0:
+            return need * DECODE_HEADROOM
+    except Exception:
+        pass
+    return 1e30
+
+
+def _resident(models):
+    """The LoadedModel entries ComfyUI currently holds for `models`.
+
+    That is the form free_memory's keep_loaded wants: it compares against the
+    entries in current_loaded_models, not against the ModelPatcher objects a node
+    is holding. Anything not matched is simply not kept, so a model that is not
+    resident costs nothing here."""
+    out = []
+    for lm in list(getattr(mm, "current_loaded_models", [])):
+        for m in models or ():
+            if m is None:
+                continue
+            try:
+                if lm.model is m or getattr(lm, "model", None) is getattr(m, "model", None):
+                    if lm not in out:
+                        out.append(lm)
+            except Exception:
+                pass
+    return out
+
+
+def _image_out_dtype():
+    """The dtype ComfyUI itself hands between nodes on THIS install.
+
+    The join used to end in a hard-coded .float(), commented "back to what every
+    downstream node expects". That was true when it was written and is not a
+    constant: ComfyUI has --fp16-intermediates, and on an install running it the
+    VAE's own decode already returns fp16 -- VAE.vae_output_dtype() IS
+    model_management.intermediate_dtype() (comfy/sd.py) -- as do EmptyLatentImage
+    and the rest of nodes.py. So on that install the node was taking frames the
+    VAE handed it in fp16, widening them to fp32 nothing had asked for, and
+    handing them to nodes whose own convention is fp16.
+
+    It is the largest thing this node holds, so the widening is not free: the
+    2580-frame chain costed at the join is 9.3GB as fp16 and 18.5GB as fp32,
+    against 44.6GB of staged weights on a 62GB machine -- which is the difference
+    between the render finishing and the OOM killer taking the server. Reported as
+    exactly that, twice.
+
+    Asked, not assumed, and never widened: whatever ComfyUI says it wants between
+    nodes is what the chain is built in. An install with the flag off is told
+    float32 and gets float32, byte for byte what it got before. Older builds have
+    no intermediate_dtype at all, so the fallback is the old constant."""
+    try:
+        return mm.intermediate_dtype()
+    except Exception:
+        return torch.float32
+
+
+def _evict_all_but(keep_model, latent=None):
+    """Unload every model EXCEPT the diffusion model from the GPU.
+
+    This is the fix for VRAM ratcheting across a long chain. soft_empty_cache()
+    only drops the CUDA allocator's cached blocks -- it does NOT unload models, so
+    ComfyUI keeps the Qwen3-VL text encoder (~14.6GB) and both VAEs resident in
+    current_loaded_models alongside the DiT. Each shot re-encodes the prompt
+    (text encoder), encodes the handoff keyframe (video VAE), then samples (DiT),
+    so all three compete for the card.
+
+    ComfyUI does free ahead of each load -- load_models_gpu() calls free_memory()
+    for what it is about to need (model_management.py:975), so the weight path is
+    not purely reactive. What it cannot size for is a long chain's ACTIVATIONS on
+    a card where the DiT is most of the VRAM. Freeing explicitly, right after
+    conditioning is built and before sampling, keeps only what the sampler needs.
+
+    ASKED FOR HONESTLY, and this is the expensive one. free_memory computes
+    `memory_to_free = memory_required - get_free_memory(device)`, so 1e30 meant
+    "unload everything but the DiT" on every shot, unconditionally -- on a 48GB card
+    with room for all of it as readily as on a 16GB one. What it unloads is the
+    ~14.6GB text encoder and both VAEs, and the next shot re-encodes the prompt and
+    the handoff keyframe, so all three come straight back. On a machine whose RAM is
+    already full of finished frames they come back from DISK, once per shot, which is
+    the thrashing this was reported as.
+
+    The DiT can size its own activations -- memory_required(shape) is what ComfyUI
+    itself calls before a load -- so ask for that. A card with room frees nothing and
+    keeps the encoder resident; a card without frees exactly as much as it must.
+    1e30 stays the fallback, because a bad estimate that frees too little turns a
+    slow render into an OOM."""
+    need = 1e30
+    try:
+        if latent is not None:
+            shape = latent["samples"].shape if isinstance(latent, dict) else latent.shape
+            need = float(keep_model.model.memory_required(tuple(shape))) * SAMPLE_HEADROOM
+            if not (need > 0):
+                need = 1e30
+    except Exception:
+        need = 1e30
+    try:
+        mm.free_memory(need, mm.get_torch_device(),
+                       keep_loaded=_resident([keep_model]))
+    except Exception:
+        try:
+            mm.soft_empty_cache(True)
+        except Exception:
+            pass
+
+
+def _sample_on_sigmas(model, seed, cfg, sampler_name, positive, negative, latent, sigmas):
+    """common_ksampler, driven by an EXTERNAL sigma schedule.
+
+    common_ksampler derives its sigmas from (sampler_name, scheduler, steps, denoise)
+    and takes no schedule argument, so a schedule computed anywhere else cannot
+    reach it. Under PDD that is fatal rather than merely inconvenient: the heads
+    accept only their nine trained boundaries, and re-deriving the grid from
+    widgets means hitting it by coincidence and losing it again the moment a step
+    count changes.
+
+    Mirrors nodes.common_ksampler's noise / mask / callback handling exactly -- the
+    only substitution is comfy.sample.sample_custom for comfy.sample.sample."""
+    latent_image = latent["samples"]
+    latent_image = comfy.sample.fix_empty_latent_channels(
+        model, latent_image,
+        latent.get("downscale_ratio_spacial", None),
+        latent.get("downscale_ratio_temporal", None))
+    noise = comfy.sample.prepare_noise(latent_image, seed, latent.get("batch_index"))
+    # `steps` here only sizes the progress bar -- the schedule is `sigmas`, whose
+    # step count is one less than its length (the trailing 0.0 is an endpoint).
+    callback = latent_preview.prepare_callback(model, max(len(sigmas) - 1, 1))
+    samples = comfy.sample.sample_custom(
+        model, noise, cfg, comfy.samplers.sampler_object(sampler_name), sigmas,
+        positive, negative, latent_image,
+        noise_mask=latent.get("noise_mask"), callback=callback,
+        disable_pbar=not comfy.utils.PROGRESS_BAR_ENABLED, seed=seed)
+    out = latent.copy()
+    out.pop("downscale_ratio_spacial", None)
+    out.pop("downscale_ratio_temporal", None)
+    out["samples"] = samples
+    return out
+
+
+RESIZE_CHUNK = 32
+
+
+def _stream_chunks(total):
+    """A collector that writes upscaled chunks into ONE destination as they land.
+
+    Both chunk loops in _upscale_frames used `out.append(...)` then
+    `frames = torch.cat(out, dim=0)`. That is the shape the finished-chain join was
+    rebuilt to stop, at a LARGER size: the list holds the whole upscaled chain and
+    the cat allocates a second one, both live at the cat, and `out` is a local that
+    is never cleared -- so it survives the cat, survives the trailing resize, and is
+    still bound at the return. Meanwhile the CALLER's pre-upscale chain cannot be
+    dropped either, because `part = frames[s:s+batch]` is a view into it.
+
+    At 2580 frames of 1056x608 that is 9.26GB per copy per doubling: 37GB x2 at 2x,
+    and 148GB x2 with the RealESRGAN_x4plus that is sitting in models/upscale_models.
+    Preallocating from the first chunk and copying into it removes exactly one of
+    those two, and drops the list at the same time.
+
+    The destination is sized from the FIRST chunk, so the model's scale factor does
+    not have to be known in advance, and the frame count is the caller's own -- an
+    upscaler changes width and height, never the number of frames."""
+    state = {"dst": None, "at": 0}
+
+    def put(piece):
+        if state["dst"] is None:
+            state["dst"] = torch.empty((int(total),) + tuple(piece.shape[1:]),
+                                       dtype=piece.dtype, device=piece.device)
+        k = int(piece.shape[0])
+        end = min(state["at"] + k, state["dst"].shape[0])
+        if end > state["at"]:
+            state["dst"][state["at"]:end].copy_(piece[:end - state["at"]])
+        state["at"] = end
+
+    def done():
+        d, at = state["dst"], state["at"]
+        if d is None:
+            return None
+        return d if at == d.shape[0] else d[:at]
+
+    return put, done
+
+
+def _resize_short_edge(frames, target, method="lanczos", chunk=0):
+    """Resize a [B,H,W,C] frame batch so its short edge == target (keeping aspect,
+    snapped to /32). Plain high-quality resize -- enlarges, doesn't add detail.
+
+    IN CHUNKS, BECAUSE LANCZOS IS FOUR FULL-LENGTH COPIES. The whole chain went
+    into one common_upscale call, and comfy.utils.lanczos is three successive list
+    comprehensions over every frame at once:
+
+        images = [Image.fromarray(...) for image in samples]        # N at source size
+        images = [image.resize(...) for image in images]            # N at target size
+        images = [torch.from_numpy(np.array(im).astype(np.float32)/255.) ...]
+        result = torch.stack(images)
+        return result.to(samples.device, samples.dtype)
+
+    A comprehension builds the new list completely before rebinding the name, so at
+    each rebind BOTH are live; then torch.stack allocates a full copy while its list
+    still exists, and .to() allocates the result while the stack still exists. Note
+    the astype(np.float32): the input is fp16 but the two largest transients are at
+    DOUBLE its width. At 2580 frames to a 1080 short edge that peaked around 147GB
+    to produce a 29GB result, and it fires on a DOWNSCALE too.
+
+    Chunked, the peak is the result plus one chunk's worth of that machinery. It is
+    bit-identical: PIL resizes each frame independently, so per-chunk and per-chain
+    give the same pixels. The early return for an already-correct size is kept, so
+    the common no-op case still allocates nothing."""
+    b, h, w, c = frames.shape
+    if min(h, w) == target:
+        return frames
+    if h <= w:
+        nh = target; nw = max(32, int(round(target * w / h / 32) * 32))
+    else:
+        nw = target; nh = max(32, int(round(target * h / w / 32) * 32))
+    step = max(1, int(chunk) or RESIZE_CHUNK)
+    out = torch.empty((b, nh, nw, c), dtype=frames.dtype, device=frames.device)
+    for i in range(0, b, step):
+        part = comfy.utils.common_upscale(
+            frames[i:i + step].movedim(-1, 1), nw, nh, method, "disabled")
+        out[i:i + step].copy_(part.movedim(1, -1))
+        del part
+    return out
+
+
+def _upscale_frames(frames, mode, model_name, target_short_edge, batch=4):
+    """Optional post-pass upscale of the finished frames (on CPU).
+      mode 'model'   : run a ComfyUI upscale model (Real-ESRGAN/UltraSharp class)
+                       via the registered loader+apply nodes, chunked with cleanup
+                       so 2000+ frames don't OOM; then fit to target short edge.
+      mode 'rtx'     : NVIDIA RTX Video Super Resolution (Tensor Cores; fastest,
+                       best quality for video -- needs Nvidia_RTX_Nodes_ComfyUI).
+      mode 'lanczos' : plain high-quality resize to the target short edge.
+    Any failure falls back to lanczos (or the raw frames), so it never breaks a
+    render. Returns (frames, note). NOTE: this SHARPENS/ENLARGES; it does not
+    reconstruct video detail the way a second-model (LTX 2.3) pass does."""
+    if mode == "off" or frames is None or getattr(frames, "shape", [0])[0] == 0:
+        return frames, ""
+    note = ""
+    if mode == "rtx":
+        # NVIDIA RTX Video Super Resolution (Comfy-Org/Nvidia_RTX_Nodes_ComfyUI).
+        # Runs on RTX Tensor Cores -- far faster than ESRGAN-class models and
+        # generally cleaner on video, though like them it enhances/enlarges rather
+        # than reconstructing detail (an LTX 2.3 re-generation does that).
+        try:
+            rtx = (_find_node(["rtx", "video", "super"]) or _find_node(["rtxvideosuperresolution"])
+                   or _find_node(["rtx", "upscale"]))
+            if rtx is None:
+                raise RuntimeError("RTX node not installed (Nvidia_RTX_Nodes_ComfyUI)")
+            scale = 2
+            if target_short_edge and int(target_short_edge) > 0:
+                cur = min(frames.shape[1], frames.shape[2])
+                if cur > 0:
+                    scale = max(1, min(4, int(round(int(target_short_edge) / cur))))
+            _put, _done = _stream_chunks(frames.shape[0])
+            n = frames.shape[0]
+            step = max(1, int(batch))
+            for st in range(0, n, step):
+                part = frames[st:st + step]
+                res = None
+                for kw in ({"image": part, "scale": scale}, {"images": part, "scale": scale},
+                           {"image": part, "scale_factor": scale}, {"image": part}):
+                    try:
+                        res = _invoke_node(rtx, **kw); break
+                    except TypeError:
+                        continue
+                if res is None:
+                    raise RuntimeError("RTX node signature not recognized")
+                _put(res.detach().to("cpu"))
+                del res, part
+                _deep_cleanup()
+            frames = _done()
+            note = f"RTX Video Super Resolution x{scale}"
+            if target_short_edge and int(target_short_edge) > 0:
+                frames = _resize_short_edge(frames, int(target_short_edge))
+                note += f"; fit to {int(target_short_edge)}px short edge"
+            return frames, note
+        except Exception as e:
+            mode = "model"
+            note = f"RTX upscale unavailable ({e}); fell back to model/lanczos"
+    if mode == "model" and model_name and model_name != "none":
+        try:
+            loader = _find_node(["upscale", "model", "load"]) or _find_node(["loadupscalemodel"])
+            applier = _find_node(["imageupscale", "model"]) or _find_node(["upscaleimageusingmodel"])
+            if loader is None or applier is None:
+                raise RuntimeError("upscale-model nodes not found")
+            up_model = _invoke_node(loader, model_name=model_name)
+            _put, _done = _stream_chunks(frames.shape[0])
+            n = frames.shape[0]
+            for s in range(0, n, max(1, int(batch))):
+                part = frames[s:s + max(1, int(batch))]
+                res = _invoke_node(applier, upscale_model=up_model, image=part)
+                _put(res.detach().to("cpu"))
+                del res, part
+                _deep_cleanup()
+            frames = _done()
+            note = f"upscaled with {model_name}"
+        except Exception as e:
+            mode = "lanczos"
+            note = f"model upscale unavailable ({e}); used lanczos"
+    if target_short_edge and int(target_short_edge) > 0:
+        try:
+            frames = _resize_short_edge(frames, int(target_short_edge))
+            note = (note + "; " if note else "") + f"fit to {int(target_short_edge)}px short edge"
+        except Exception as e:
+            note = (note + "; " if note else "") + f"resize failed ({e})"
+    elif mode == "lanczos" and not note:
+        note = "lanczos selected but no target set -> unchanged"
+    return frames, note
+
+
+def _find_node(substrings):
+    """Find a registered node whose key contains all of `substrings` (lowercased)."""
+    maps = getattr(nodes, "NODE_CLASS_MAPPINGS", {}) or {}
+    for k, v in maps.items():
+        kl = k.lower()
+        if all(s in kl for s in substrings):
+            return v
+    return None
+
+
+def _invoke_node(cls, **kwargs):
+    """Call a registered ComfyUI node (V1 FUNCTION or V3 execute) with kwargs and
+    return its first output. Used to reuse ComfyUI's own upscale-model loader/apply
+    so we don't reimplement spandrel loading or tiled scaling."""
+    inst = cls()
+    fn = None
+    if getattr(cls, "FUNCTION", None) and hasattr(inst, cls.FUNCTION):
+        fn = getattr(inst, cls.FUNCTION)
+    else:
+        for cand in ("execute", "upscale", "load_model", "load"):
+            if hasattr(inst, cand):
+                fn = getattr(inst, cand); break
+    if fn is None:
+        raise RuntimeError("no callable entrypoint")
+    out = fn(**kwargs)
+    out = getattr(out, "result", out)
+    return out[0] if isinstance(out, (tuple, list)) else out
+
+
+# --- THE GRADE THE CHAIN ADDS TO ITSELF -------------------------------------
+# Every shot boundary decodes a shot, hands its LAST frame over, and re-encodes that as
+# the next shot's keyframe. The distill reproduces the keyframe faithfully enough to
+# inherit whatever is already in it and SYNTHESISES frame 0 rather than copying it, so
+# its own bias lands on top: S_next = a*S + b, a near 1, b above 0. Linear at best,
+# geometric at worst, invisible shot to shot. And the VAE hard-clips every decode to
+# 0..1, which makes the expansion a RATCHET -- headroom spent is not recoverable, so it
+# shows as crushed blacks and blown highlights rather than merely as more contrast.
+#
+# These two are the measurement and the correction. Both work per colour channel,
+# because the clip is per channel: the VAE un-whitens with ImageNet stds before it
+# clamps, so the 0..1 rails sit at different distances in each channel and the blue
+# floor and red ceiling bite first. A single luma number would miss the colour half.
+LEVEL_POOL = 64                # cells per axis the level statistics are measured on
+
+
+def frame_levels(img):
+    """(mean, std) per colour channel for one frame, as 3-vectors, or (None, None).
+
+    Area-pooled to LEVEL_POOL first, so a pre-upscale frame and an upscaled one can be
+    compared: pooling measures the PICTURE's levels rather than its resolution. Measured
+    across a 2x resize, std agrees to 0.28% on picture-like content -- and to only 15%
+    on pure noise, because pooling cannot preserve variance that lives entirely at the
+    pixel scale. Real frames are the former, and whatever residual there is cancels
+    anyway: the caller measures the same pipeline difference separately and subtracts it.
+
+    float32 throughout, deliberately: these frames are fp16 under
+    --fp16-intermediates, and an fp16 mean accumulated over a 1344x768 frame biases
+    badly enough to matter at the sizes being corrected here."""
+    x = img
+    if x.dim() == 4:
+        x = x[0]
+    if x.dim() != 3 or int(x.shape[-1]) < 3:
+        return None, None
+    if int(x.shape[0]) < 2 or int(x.shape[1]) < 2:
+        return None, None
+    x = x[..., :3].float().permute(2, 0, 1).unsqueeze(0)
+    p = torch.nn.functional.adaptive_avg_pool2d(x, LEVEL_POOL)[0].reshape(3, -1)
+    return p.mean(dim=1), p.std(dim=1)
+
+
+# The per-shot motion envelope lived here, measured so the built footsteps could be
+# timed off the picture. Nothing is built any more -- see the note at the top of
+# audio.py -- so there is nothing left to time, and a measurement with no reader is a
+# measurement that rots. Removed with the synthesiser it served.
+
+
+def apply_levels(img, gain, offset):
+    """Rescale a frame's contrast and level about its OWN per-channel mean.
+
+    The pivot is the frame's own mean and never a target. That is the whole reason this
+    can run on any scene: a beat that walks into a darker room keeps its darkness,
+    because nothing here knows or cares what the level is -- only how much the last
+    boundary expanded it. Anchoring to shot 1 instead would cancel every deliberate
+    lighting change in the film, which is the opposite failure.
+
+    Clamped into 0..1 because the next thing that happens to this frame is an 8-bit
+    quantisation (comfy.utils.common_upscale goes through a uint8 PIL round trip even
+    at the same size), so there is no headroom outside the range to borrow from."""
+    x = img.float()
+    c = min(3, int(x.shape[-1]))
+    m = x[..., :c].reshape(-1, c).mean(dim=0)
+    g = gain[:c].to(device=x.device, dtype=x.dtype)
+    o = offset[:c].to(device=x.device, dtype=x.dtype)
+    y = x.clone()
+    y[..., :c] = ((x[..., :c] - m) * g + m + o).clamp(0.0, 1.0)
+    return y.to(img.dtype)

dumas_h3_longvideos_shot_plan.py

@@ -0,0 +1,109 @@
+# H3-LongVideos -- https://github.com/Smite79/MiniMax-H3-LongVideos
+# Copyright (c) 2026 Smite79. All rights reserved.
+# Redistribution, in whole or in part, requires written permission.
+# This notice may not be removed or altered. See LICENSE.
+"""Per-shot records passed from prompt planning to rendering."""
+
+from dataclasses import dataclass, field
+
+
+@dataclass
+class Shot:
+    prompt: str
+    cast: list[str]
+    speech: bool
+    sounded: bool
+    voiced_only: bool
+    events: list[str]
+    frame_count: int = 0
+    refs: list[object] = field(default_factory=list)
+    line_seconds: float = 0.0     # the planner's estimate of the spoken line, words / WORDS_PER_SEC
+
+
+@dataclass
+class ShotPlan:
+    shots: list[Shot] = field(default_factory=list)
+
+    @property
+    def prompts(self):
+        return [shot.prompt for shot in self.shots]
+
+    def add(self, prompt, cast, speech, sounded, voiced_only, events):
+        shot = Shot(prompt, list(cast or ()), bool(speech), bool(sounded),
+                    bool(voiced_only), list(events or ()))
+        self.shots.append(shot)
+
+    def set_frame_counts(self, counts):
+        counts = [int(n) for n in counts]
+        if len(counts) != len(self.shots) or any(n <= 0 for n in counts):
+            raise ValueError("frame counts must be positive and match the planned shots")
+        for shot, count in zip(self.shots, counts):
+            shot.frame_count = count
+
+    def validate(self):
+        if any(shot.frame_count <= 0 for shot in self.shots):
+            raise ValueError("each shot needs a positive frame count before rendering")
+        return self
+
+    def __len__(self):
+        return len(self.shots)
+
+
+@dataclass
+class PreparedVideo:
+    """Resolved inputs consumed by the render stage; model/tensor handles are shared."""
+    _placed_shots: object
+    _first_is_plate: object
+    _returns: object
+    _soft_landing: object
+    _tagged_names: object
+    ambient_audio: object
+    ambient_level: float
+    apply_model_sampling: bool
+    audio_vae: object
+    auto_sound: bool
+    bared_shots: object
+    cfg: float
+    cleanup_between_shots: bool
+    clip: object
+    first_frame: object
+    foley_level: float
+    h: int
+    latent_upscale: str
+    latent_upscale_scale: float
+    megapixels: float
+    model: object
+    moved_shots: object
+    negative: object
+    notes: list[str]
+    plan: ShotPlan
+    ref_noise_aug: float | None
+    restart_after_removal: bool
+    revealed_shots: object
+    sampler_name: str
+    scheduler: str
+    seed: int
+    shift_audio: float
+    shift_video: float
+    sigmas: object
+    silence_nonspeech: bool
+    speech_lead_seconds: float
+    speech_tail_seconds: float
+    hold_levels: float
+    handoff_frames: int
+    staging_shots: object
+    steps: int
+    stripped_shots: object
+    cut_shots: object
+    tiled_decode: bool
+    trim_seam: bool
+    upscale: str
+    upscale_batch: int
+    upscale_model: str
+    upscale_target_short_edge: int
+    vae: object
+    w: int
+    shot_rooms: object = None         # {0-based shot: (room it opens in, room it ends in)}
+    hardware_changed: object = None   # 1-based shots that put hardware on or take it off
+    shot_frames: object = None        # {0-based shot: (who its frames show, who is still there at its end)}
+    reentry_shots: object = None      # {0-based shot: who walks in while the keyframe still has them}

js/h3_beat_prompt/index.js

@@ -8,32 +8,15 @@ const DEFAULT_W = 520;
 const DEFAULT_H = 340;
 const DEFAULT_BEAT = "Describe this beat.";
 const STATE_PROPERTY = "dumas_h3_beat_prompt_state";
-const CONTINUITY_OPTIONS = ["", "soft carry", "hard cut", "keyframe carry", "handoff ref"];
-const REF_MODE_OPTIONS = ["", "auto ref2v", "where tagged", "first shot", "every shot", "every shot + handoff ref"];
 const MANAGED_DIRECTIVES = {
-  seconds: ["seconds", "duration"],
-  continuity: ["continuity"],
-  ref_mode: ["ref_mode"],
-  ref_noise_aug: ["ref_noise_aug"],
-  anchor_add: ["anchor_add"],
-  overall_soundscape: ["overall_soundscape", "soundscape"],
-  non_diegetic_music: ["non_diegetic_music", "music"],
+  remove: ["remove", "removed", "off"],
+  add: ["add", "wear", "wearing"],
 };
 const DIRECTIVE_EXAMPLES = [
-  ["wardrobe set", "wardrobe: Maya = grey shorts, red jacket"],
-  ["wardrobe add", "wardrobe: Maya += red jacket"],
-  ["wardrobe remove", "wardrobe: Maya -= red jacket"],
-  ["seconds", "seconds: 8"],
-  ["exit", "exit: Maya"],
-  ["enter", "enter: Jon"],
-  ["continuity", "continuity: hard cut"],
-  ["ref_mode", "ref_mode: every shot"],
-  ["ref_noise_aug", "ref_noise_aug: 0.92"],
-  ["anchor_add", "anchor_add: harsh sodium-vapor spill, wet pavement, long-lens compression"],
-  ["overall_soundscape", "overall_soundscape: soft rain, distant traffic"],
-  ["non_diegetic_music", "non_diegetic_music: tense analog synth pulse"],
-  ["soundscape", "soundscape: fluorescent room tone, faint HVAC hum"],
-  ["music", "music: low ominous cello and sparse percussion"],
+  ["remove", "remove: red jacket"],
+  ["off", "off: steel collar"],
+  ["add", "add: white shirt underneath"],
+  ["wearing", "wearing: black coat"],
 ];
 
 function injectCSS() {
@@ -183,7 +166,7 @@ function injectCSS() {
 }
 
 function defaultState() {
-  return { beats: [{ text: DEFAULT_BEAT }] };
+  return { scene: "", character_sheet: "", beats: [{ text: DEFAULT_BEAT }] };
 }
 
 function normalizeState(value) {
@@ -200,7 +183,11 @@ function normalizeState(value) {
   const normalized = beats.map((beat) => ({
     text: typeof beat?.text === "string" ? beat.text : String(beat?.text || ""),
   }));
-  return normalized.length ? { beats: normalized } : defaultState();
+  return {
+    scene: typeof parsed.scene === "string" ? parsed.scene : String(parsed.scene || ""),
+    character_sheet: typeof parsed.character_sheet === "string" ? parsed.character_sheet : String(parsed.character_sheet || ""),
+    beats: normalized.length ? normalized : [{ text: DEFAULT_BEAT }],
+  };
 }
 
 function readState(node) {
@@ -321,6 +308,51 @@ function renderUI(node) {
   node._dh3bpRenderedState = JSON.stringify(state);
   ui.list.innerHTML = "";
 
+  const buildTopTextarea = ({ labelText, placeholder, value, onInput }) => {
+    const card = document.createElement("div");
+    card.className = "dh3bp-beat";
+
+    const label = document.createElement("div");
+    label.className = "dh3bp-label";
+    label.textContent = labelText;
+
+    const textarea = document.createElement("textarea");
+    textarea.className = "dh3bp-text";
+    textarea.placeholder = placeholder;
+    textarea.value = value || "";
+    textarea.addEventListener("input", () => {
+      onInput(textarea.value);
+      updateTextareaHeight(textarea);
+    });
+    textarea.addEventListener("keydown", stopCanvasKeyboard);
+
+    card.append(label, textarea);
+    updateTextareaHeight(textarea);
+    return card;
+  };
+
+  ui.list.appendChild(buildTopTextarea({
+    labelText: "Scene paragraph",
+    placeholder: "Optional. Persistent location, lighting, camera, tone. Leave empty if you wire the Long Videos anchor input.",
+    value: state.scene,
+    onInput: (value) => {
+      const next = readState(node);
+      next.scene = value;
+      writeState(node, next);
+    },
+  }));
+
+  ui.list.appendChild(buildTopTextarea({
+    labelText: "Character sheet",
+    placeholder: "Optional. One character per line, e.g. Maya: 27, she, silver hair, red jacket, the woman in <Picture 1>.",
+    value: state.character_sheet,
+    onInput: (value) => {
+      const next = readState(node);
+      next.character_sheet = value;
+      writeState(node, next);
+    },
+  }));
+
   state.beats.forEach((beat, index) => {
     const card = document.createElement("div");
     card.className = "dh3bp-beat";
@@ -379,85 +411,27 @@ function renderUI(node) {
       return wrap;
     };
 
-    const secondsInput = document.createElement("input");
-    secondsInput.className = "dh3bp-input";
-    secondsInput.type = "text";
-    secondsInput.placeholder = "8";
-    secondsInput.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.seconds);
-    secondsInput.addEventListener("input", () => {
-      applyTextUpdate(setDirectiveValue(textarea.value, "seconds", MANAGED_DIRECTIVES.seconds, secondsInput.value));
+    const removeInput = document.createElement("input");
+    removeInput.className = "dh3bp-input";
+    removeInput.type = "text";
+    removeInput.placeholder = "red jacket";
+    removeInput.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.remove);
+    removeInput.addEventListener("input", () => {
+      applyTextUpdate(setDirectiveValue(textarea.value, "remove", MANAGED_DIRECTIVES.remove, removeInput.value));
     });
 
-    const continuitySelect = document.createElement("select");
-    continuitySelect.className = "dh3bp-select";
-    CONTINUITY_OPTIONS.forEach((value) => {
-      const option = document.createElement("option");
-      option.value = value;
-      option.textContent = value || "Default";
-      continuitySelect.appendChild(option);
-    });
-    continuitySelect.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.continuity);
-    continuitySelect.addEventListener("change", () => {
-      applyTextUpdate(setDirectiveValue(textarea.value, "continuity", MANAGED_DIRECTIVES.continuity, continuitySelect.value));
-    });
-
-    const refModeSelect = document.createElement("select");
-    refModeSelect.className = "dh3bp-select";
-    REF_MODE_OPTIONS.forEach((value) => {
-      const option = document.createElement("option");
-      option.value = value;
-      option.textContent = value || "Global";
-      refModeSelect.appendChild(option);
-    });
-    refModeSelect.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.ref_mode);
-    refModeSelect.addEventListener("change", () => {
-      applyTextUpdate(setDirectiveValue(textarea.value, "ref_mode", MANAGED_DIRECTIVES.ref_mode, refModeSelect.value));
-    });
-
-    const refNoiseInput = document.createElement("input");
-    refNoiseInput.className = "dh3bp-input";
-    refNoiseInput.type = "text";
-    refNoiseInput.placeholder = "0.95";
-    refNoiseInput.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.ref_noise_aug);
-    refNoiseInput.addEventListener("input", () => {
-      applyTextUpdate(setDirectiveValue(textarea.value, "ref_noise_aug", MANAGED_DIRECTIVES.ref_noise_aug, refNoiseInput.value));
-    });
-
-    const anchorInput = document.createElement("input");
-    anchorInput.className = "dh3bp-input";
-    anchorInput.type = "text";
-    anchorInput.placeholder = "extra per-shot style treatment";
-    anchorInput.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.anchor_add);
-    anchorInput.addEventListener("input", () => {
-      applyTextUpdate(setDirectiveValue(textarea.value, "anchor_add", MANAGED_DIRECTIVES.anchor_add, anchorInput.value));
-    });
-
-    const soundscapeInput = document.createElement("input");
-    soundscapeInput.className = "dh3bp-input";
-    soundscapeInput.type = "text";
-    soundscapeInput.placeholder = "faint traffic, loose sign rattle";
-    soundscapeInput.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.overall_soundscape);
-    soundscapeInput.addEventListener("input", () => {
-      applyTextUpdate(setDirectiveValue(textarea.value, "overall_soundscape", MANAGED_DIRECTIVES.overall_soundscape, soundscapeInput.value));
-    });
-
-    const musicInput = document.createElement("input");
-    musicInput.className = "dh3bp-input";
-    musicInput.type = "text";
-    musicInput.placeholder = "low pulsing synth tension";
-    musicInput.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.non_diegetic_music);
-    musicInput.addEventListener("input", () => {
-      applyTextUpdate(setDirectiveValue(textarea.value, "non_diegetic_music", MANAGED_DIRECTIVES.non_diegetic_music, musicInput.value));
+    const addInput = document.createElement("input");
+    addInput.className = "dh3bp-input";
+    addInput.type = "text";
+    addInput.placeholder = "white shirt underneath";
+    addInput.value = readDirectiveValue(beat.text, MANAGED_DIRECTIVES.add);
+    addInput.addEventListener("input", () => {
+      applyTextUpdate(setDirectiveValue(textarea.value, "add", MANAGED_DIRECTIVES.add, addInput.value));
     });
 
     controls.append(
-      buildField({ labelText: "Seconds", input: secondsInput }),
-      buildField({ labelText: "Continuity", input: continuitySelect }),
-      buildField({ labelText: "Ref Mode", input: refModeSelect }),
-      buildField({ labelText: "Ref Noise Aug", input: refNoiseInput }),
-      buildField({ labelText: "Anchor Add", className: "dh3bp-control-wide", input: anchorInput }),
-      buildField({ labelText: "Shot Soundscape", className: "dh3bp-control-wide", input: soundscapeInput }),
-      buildField({ labelText: "Shot Music", className: "dh3bp-control-wide", input: musicInput }),
+      buildField({ labelText: "Remove from memory", input: removeInput }),
+      buildField({ labelText: "Add to memory", input: addInput }),
     );
 
     const directives = document.createElement("div");
@@ -501,7 +475,7 @@ function setupNode(node) {
   title.textContent = "Beat Prompt Builder";
   const subtitle = document.createElement("div");
   subtitle.className = "dh3bp-subtitle";
-  subtitle.textContent = "One textbox per H3 beat, plus per-shot controls for timing, ref behavior, continuity, anchor adds, and audio directives.";
+  subtitle.textContent = "Upstream Long Videos format: optional scene, optional character sheet, then one blank-line-separated beat per shot.";
   titleWrap.append(title, subtitle);
 
   const addButton = document.createElement("button");

tests/test_dumas_h3_beat_prompt.py

@@ -11,35 +11,62 @@ class DumasH3BeatPromptTests(unittest.TestCase):
         state = self.module._parse_beat_prompt_state("not json")
         self.assertEqual(
             state,
-            {"beats": [{"text": "Describe this beat."}]},
+            {
+                "scene": "",
+                "character_sheet": "",
+                "beats": [{"text": "Describe this beat."}],
+            },
         )
 
-    def test_assemble_prompt_joins_beats_with_blank_lines(self):
+    def test_assemble_prompt_outputs_upstream_sections(self):
         prompt = self.module._assemble_beat_prompt(
             {
+                "scene": "A rainy kitchen at night.",
+                "character_sheet": "Maya: 27, she, red jacket, silver hair.",
                 "beats": [
-                    {"text": "A woman enters the room."},
-                    {"text": "wardrobe: Maya = red jacket\nShe sits at the table."},
+                    {"text": "Maya enters the room."},
+                    {"text": "remove: red jacket\nadd: white shirt underneath\nShe sits at the table."},
                     {"text": "  "},
-                    {"text": "music: low synth pulse"},
                 ]
             }
         )
         self.assertEqual(
             prompt,
             (
-                "A woman enters the room.\n\n"
-                "wardrobe: Maya = red jacket\nShe sits at the table.\n\n"
-                "music: low synth pulse"
+                "A rainy kitchen at night.\n\n"
+                "Maya: 27, she, red jacket, silver hair.\n\n"
+                "Maya enters the room.\n\n"
+                "remove: red jacket\nadd: white shirt underneath\nShe sits at the table."
             ),
         )
 
+    def test_assemble_prompt_strips_old_dumas_directives(self):
+        prompt = self.module._assemble_beat_prompt(
+            {
+                "beats": [
+                    {
+                        "text": (
+                            "seconds: 8\n"
+                            "continuity: hard cut\n"
+                            "ref_mode: every shot\n"
+                            "soundscape: soft rain\n"
+                            "music: low synth\n"
+                            "Maya opens the cupboard.\n"
+                            "remove: red jacket"
+                        )
+                    },
+                ]
+            }
+        )
+
+        self.assertEqual(prompt, "Maya opens the cupboard.\nremove: red jacket")
+
     def test_node_build_prompt_uses_hidden_state(self):
         node = self.module.DumasH3BeatPromptNode()
         result = node.build_prompt(
-            '{"beats":[{"text":"Beat one"},{"text":"Beat two"}]}'
+            '{"scene":"Scene","character_sheet":"Maya: 27, she","beats":[{"text":"Beat one"},{"text":"Beat two"}]}'
         )
-        self.assertEqual(result, ("Beat one\n\nBeat two",))
+        self.assertEqual(result, ("Scene\n\nMaya: 27, she\n\nBeat one\n\nBeat two",))
 
 
 if __name__ == "__main__":

tests/test_dumas_h3_longvideos.py

@@ -23,6 +23,16 @@ class DumasH3LongVideosUpstreamWrapperTests(unittest.TestCase):
                 "folder_paths",
                 "dumas_h3_longvideos",
                 "dumas_h3_longvideos_upstream",
+                "dumas_h3_longvideos_engine",
+                "dumas_h3_longvideos_shot_plan",
+                "dumas_h3_longvideos_runtime",
+                "dumas_h3_longvideos_audio",
+                "dumas_h3_longvideos_conditioning",
+                "h3_engine",
+                "h3_shot_plan",
+                "h3_runtime",
+                "h3_audio",
+                "h3_conditioning",
             )
         }
 
@@ -112,8 +122,20 @@ class DumasH3LongVideosUpstreamWrapperTests(unittest.TestCase):
         self.assertIn("first_frame", schema["optional"])
         self.assertIn("ref_image_1", schema["optional"])
         self.assertIn("latent_upscale", schema["optional"])
+        self.assertIn("speech_tail_seconds", schema["optional"])
+        self.assertIn("hold_camera", schema["optional"])
+        self.assertIn("verbatim", schema["optional"])
+        self.assertIn("handoff_frames", schema["optional"])
+        self.assertEqual(schema["optional"]["handoff_frames"][1]["default"], 1)
         self.assertEqual(node_cls.RETURN_NAMES[0:4], ("images", "audio", "info", "script"))
 
+    def test_handoff_context_claim_names_reference_range(self):
+        upstream = importlib.import_module("dumas_h3_longvideos_upstream")
+
+        self.assertIn("<Picture 2> through <Picture 22>", upstream.handoff_context_claim(2, 22))
+        self.assertIn("no new subjects", upstream.handoff_context_claim(2, 22))
+        self.assertIn("<Picture 5>", upstream.handoff_context_claim(5, 5))
+
 
 class _NullContext:
     def __enter__(self):