Expand latent upscale spatial stitch controls

H3_LONG_VIDEOS_GUIDE.md

@@ -1164,6 +1164,7 @@ What this really means:
 - the conditioning is rebuilt at that target size
 - the node then runs a short refinement pass over the upscaled latent with the sampler, scheduler, step count, denoise, and megapixel target you picked on the latent-upscale params node
 - if the target is larger than the spatial tile size, that refinement pass is processed in spatial batches using the same tile defaults as the upstream latent-split node
+- the spatial stitch mode follows the upstream overlap controls, including `linear`, `smoothstep`, `overwrite`, and `midpoint`
 
 Good starting point:
 
@@ -1172,6 +1173,7 @@ Good starting point:
 - start with `euler_ancestral`, `simple`, `2` steps, and `0.2` denoise
 - leave width and height at `0` unless you want an exact override; otherwise `megapixels` drives the target size
 - leave the spatial tile inputs at their defaults first: `512x512` tiles, `64` overlap, `0` fade width, `earlier` overlap mode
+- keep `linear` blend first unless you want to reproduce a specific upstream stitch style
 
 The important part is that this stage is still a latent pass, not a pixel-space resize:
 

README.md

@@ -47,7 +47,7 @@
   - Per-shot directives now support `continuity:`, `ref_mode:`, `ref_noise_aug:`, `anchor_add:`, `soundscape:`, and `music:` in addition to the existing timing and wardrobe directives.
 
 - `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`, `overlap_mode`
+  - Inputs: `mode`, `model_name`, `method`, `width`, `height`, `device`, `precision`, `sampler_name`, `scheduler`, `steps`, `denoise`, `megapixels`, `tile_width`, `tile_height`, `overlap`, `fade_width`, `overlap_mode`, `overlap_blend`
   - 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 optional spatial batching.
 

dumas_h3_latent_upscale.py

@@ -473,10 +473,12 @@ class H3LatentUpscaleParams:
                     "tooltip": "Width in pixels of the freeze-to-free transition inside each overlap strip. 0 freezes the whole strip, matching the upstream default."}),
                 "overlap_mode": (["earlier", "later"], {"default": "earlier",
                     "tooltip": "Which tile wins the overlap band when stitching the spatial batches back together."}),
+                "overlap_blend": (["linear", "smoothstep", "overwrite", "midpoint"], {"default": "linear",
+                    "tooltip": "How overlap bands are blended when the spatial batches are stitched back together."}),
             }
         }
 
-    def build(self, mode, model_name, method, width, height, device, precision, sampler_name, scheduler, steps, denoise, megapixels, tile_width, tile_height, overlap, fade_width, overlap_mode):
+    def build(self, mode, model_name, method, width, height, device, precision, sampler_name, scheduler, steps, denoise, megapixels, tile_width, tile_height, overlap, fade_width, overlap_mode, overlap_blend):
         width = int(width)
         height = int(height)
         steps = int(steps)
@@ -500,6 +502,7 @@ class H3LatentUpscaleParams:
                 "overlap": overlap,
                 "fade_width": fade_width,
                 "overlap_mode": overlap_mode,
+                "overlap_blend": overlap_blend,
             },)
         if width > 0:
             width = int(round(width / 32.0)) * 32
@@ -524,6 +527,7 @@ class H3LatentUpscaleParams:
             "overlap": overlap,
             "fade_width": fade_width,
             "overlap_mode": overlap_mode,
+            "overlap_blend": overlap_blend,
         },)
 
 

dumas_h3_longvideos.py

@@ -4104,10 +4104,18 @@ def _latent_spatial_grid(h, w, th, tw, ol_h, ol_w):
     return rows, cols, trows, tcols
 
 
-def _latent_spatial_blend_weights(t, overlap_mode):
+def _latent_spatial_blend_weights(t, overlap_mode, overlap_blend="linear"):
+    if overlap_blend == "overwrite":
+        return torch.ones_like(t) if overlap_mode == "later" else torch.zeros_like(t)
+    if overlap_blend == "midpoint":
+        base = (t >= 0.5).to(t.dtype)
+    elif overlap_blend == "smoothstep":
+        base = t * t * (3.0 - 2.0 * t)
+    else:
+        base = t
     if overlap_mode == "later":
-        return 1.0 - t
-    return t
+        return base
+    return 1.0 - base
 
 
 def _nested_tensor_parts(samples):
@@ -6510,6 +6518,7 @@ class H3LongVideos:
                 overlap_px = max(0, int(latent_upscale_param.get("overlap", 64) or 64))
                 fade_px = max(0, int(latent_upscale_param.get("fade_width", 0) or 0))
                 overlap_mode = str(latent_upscale_param.get("overlap_mode", "earlier"))
+                overlap_blend = str(latent_upscale_param.get("overlap_blend", "linear"))
                 tile_tw = max(1, min(int(up_w), max(1, tile_w_px // 16)))
                 tile_th = max(1, min(int(up_h), max(1, tile_h_px // 16)))
                 ol_tw = max(0, min(tile_tw - 1, overlap_px // 16))
@@ -6548,7 +6557,7 @@ class H3LongVideos:
                             region.copy_(tile_video_out)
                             if col_index > 0 and ol_tw > 0:
                                 t = torch.linspace(0.0, 1.0, ol_tw, device=region.device, dtype=region.dtype)
-                                w = _latent_spatial_blend_weights(t, overlap_mode)
+                                w = _latent_spatial_blend_weights(t, overlap_mode, overlap_blend)
                                 if fw_tw > 0:
                                     w = w.clone()
                                     w[:fw_tw] = 0.0
@@ -6558,7 +6567,7 @@ class H3LongVideos:
                                 )
                             if row_index > 0 and ol_th > 0:
                                 t = torch.linspace(0.0, 1.0, ol_th, device=region.device, dtype=region.dtype)
-                                w = _latent_spatial_blend_weights(t, overlap_mode)
+                                w = _latent_spatial_blend_weights(t, overlap_mode, overlap_blend)
                                 if fw_th > 0:
                                     w = w.clone()
                                     w[:fw_th] = 0.0
@@ -6707,8 +6716,9 @@ class H3LongVideos:
             tile_w_px = int(latent_upscale_param.get("tile_width", 512) or 512)
             tile_h_px = int(latent_upscale_param.get("tile_height", 512) or 512)
             overlap_px = max(0, int(latent_upscale_param.get("overlap", 64) or 64))
+            overlap_blend = str(latent_upscale_param.get("overlap_blend", "linear"))
             if tile_w_px > 0 and tile_h_px > 0 and (tile_w_px < target_w or tile_h_px < target_h):
-                batch_note = f"; spatial batches {tile_w_px}x{tile_h_px}px overlap {overlap_px}px"
+                batch_note = f"; spatial batches {tile_w_px}x{tile_h_px}px overlap {overlap_px}px {overlap_blend}"
             latent_upscale_note = (
                 f" latent upscale: target {target_w}x{target_h}px{detail}; "
                 f"{int(latent_upscale_param.get('steps', 2) or 2)}-step refinement "

tests/test_dumas_h3_longvideos.py

@@ -1132,6 +1132,7 @@ class DumasH3LongVideosHelperTests(unittest.TestCase):
         self.assertEqual(required["overlap"][1]["default"], 64)
         self.assertEqual(required["fade_width"][1]["default"], 0)
         self.assertEqual(required["overlap_mode"][1]["default"], "earlier")
+        self.assertEqual(required["overlap_blend"][1]["default"], "linear")
 
     def test_compose_persistent_does_not_expand_ambiguous_plural_to_full_cast(self):
         active = self.module.parse_wardrobe(