README.md

README.md

@@ -8,6 +8,10 @@ All of my nodes are created with the help of AI, so there may or may not be redu
 
 ## ▶️ YouTube Tutorial Videos
 
+<a href="https://youtu.be/j28z5PZXkKk?si=TPeWKm2BW-uA3BAx">
+<img width="642" height="718" alt="Capture d&#39;écran 2026-06-17 234136" src="https://github.com/user-attachments/assets/89433110-8b78-4971-8098-3a9e9226f460" />
+
+
 <table>
   <tr>
     <td>
@@ -28,13 +32,31 @@ All of my nodes are created with the help of AI, so there may or may not be redu
 ## ❓ How to install nodes
 
 - Navigate to your `/ComfyUI/custom_nodes/ folder`
-- Run `git clone https://github.com/WhatDreamscost/WhatDreamsCost-ComfyUI`
+- Delete your old "WhatDreamsCost-ComfyUI" Folder
-- Or download through the ComfyUI Manager.
+- Run `git clone -b main_cs https://github.com/CGlide/WhatDreamsCost-ComfyUI.git`
+
 
 **❗❗IMPORTANT❗❗**
 
-If you don't see the latest version (v1.3.5) yet in the manager then just downloaded the nightly version (or fetch the updates to update the list to see the latest version). 
+
-Also you will need to update ComfyUI-LTXVideo and ComfyUI-KJNodes to the latest version as well. You cannot use this node without updating ComfyUI-LTXVideo!
+This is a Modded LTX director node with some extra options to help you create videos with references sheets
+
+This node uses Ollama Locally : you will need to install it (very lightweight) and 
+
+- In your Ollama model folder run this command to install qwen 3.5 2b q4 (1.9gb) "ollama run huihui_ai/qwen3.5-abliterated:2B"
+
+- The model won't eat memory while generating since there is an auto clear VRAM when you hit Run or after 5min.
+
+- You can still enter your description manually if you don't want to install it but it works very well!
+
+- Both references mode work (fixed), Licon MSR and Ghost Mask.
+
+- Use their Lora, that is an important step : https://huggingface.co/LiconStudio/LTX-2.3-Multiple-Subject-Reference/tree/main
+
+- Enjoy!
+
+<img width="596" height="668" alt="Capture d&#39;écran 2026-06-13 234036" src="https://github.com/user-attachments/assets/2dfe7992-a72a-4eda-a103-41b3a8f77714" />
+
 
 # 🔄 Recent Updates
 

ltx_director_guide.py

@@ -24,6 +24,7 @@ class LTXDirectorGuide(LTXVAddGuide):
                 GuideData.Input("guide_data", tooltip="Guide data produced by Prompt Relay Encode (Timeline)."),
                 io.Float.Input("scale_by", default=1.0, min=0.01, max=8.0, step=0.01, tooltip="Scale the latent by this factor."),
                 io.Combo.Input("upscale_method", options=["nearest-exact", "bilinear", "area", "bicubic", "bislerp"], default="bicubic", tooltip="Method used to upscale/downscale the latent."),
+                io.Float.Input("msr_strength", default=0.0, min=0.0, max=1.0, step=0.05, tooltip="Licon MSR only: per-stage reference strength override. 0 = use the Director's value (full pull, right for stage 1). On a refinement/upscale stage set ~0.4 to hold detail without the references repainting the opening (fixes stage-2 mist/ghosting)."),
             ],
             outputs=[
                 io.Conditioning.Output(display_name="positive"),
@@ -33,7 +34,7 @@ class LTXDirectorGuide(LTXVAddGuide):
         )
 
     @classmethod
-    def execute(cls, positive, negative, vae, latent, guide_data, scale_by=1.0, upscale_method="bicubic") -> io.NodeOutput:
+    def execute(cls, positive, negative, vae, latent, guide_data, scale_by=1.0, upscale_method="bicubic", msr_strength=0.0) -> io.NodeOutput:
         scale_factors = vae.downscale_index_formula
 
         # Clone latents to avoid mutating upstream nodes
@@ -68,6 +69,12 @@ class LTXDirectorGuide(LTXVAddGuide):
 
         _, _, latent_length, latent_height, latent_width = latent_image.shape
 
+        # MSR mode: the Director handed us the raw references; inject them at THIS node's
+        # resolution (so a half-res Stage 1 and a full-res Stage 2 each stay self-consistent).
+        msr = guide_data.get("msr")
+        if msr is not None:
+            return cls._inject_msr(positive, negative, vae, latent_image, noise_mask, msr, msr_strength)
+
         images = guide_data.get("images", [])
         insert_frames = guide_data.get("insert_frames", [])
         strengths = guide_data.get("strengths", [])
@@ -88,3 +95,60 @@ class LTXDirectorGuide(LTXVAddGuide):
             )
 
         return io.NodeOutput(positive, negative, {"samples": latent_image, "noise_mask": noise_mask})
+
+    @classmethod
+    def _inject_msr(cls, positive, negative, vae, latent_image, noise_mask, msr, msr_strength=0.0):
+        """Inject the Licon MSR references at this node's current resolution."""
+        from nodes import NODE_CLASS_MAPPINGS
+        from .ltx_director import _execute_comfy_node, _unpack
+
+        IC = NODE_CLASS_MAPPINGS.get("LTXAddVideoICLoRAGuide")
+        AG = NODE_CLASS_MAPPINGS.get("LTXVAddGuide")
+        if IC is None or AG is None:
+            raise ValueError("MSR mode needs LTXAddVideoICLoRAGuide (ComfyUI-LTXVideo) and LTXVAddGuide (core LTXV nodes).")
+
+        def clamp01(v):
+            return max(0.0, min(1.0, float(v)))
+
+        slideshow = msr["slideshow"]
+        keyframes = msr["keyframes"]
+        prefix_latents = int(msr["prefix_latents"])
+        # Per-stage override: msr_strength > 0 wins (set it low on a refinement stage so the
+        # references hold detail without repainting the opening); 0 = use the Director's value.
+        strength = clamp01(msr_strength) if float(msr_strength) > 0.0 else clamp01(msr["strength"])
+        downscale = float(msr["downscale"])
+
+        # Pad the generated region so (pad + keyframes) == prefix_latents, so the downstream crop
+        # (which trims the slideshow's temporal footprint) lands on the true clean length.
+        pad_latents = max(0, prefix_latents - len(keyframes))
+        if pad_latents > 0:
+            B, C, F, H, W = latent_image.shape
+            latent_image = torch.cat(
+                [latent_image, torch.zeros((B, C, pad_latents, H, W), dtype=latent_image.dtype, device=latent_image.device)],
+                dim=2,
+            )
+            mb, mc, mf, mh, mw = noise_mask.shape
+            noise_mask = torch.cat(
+                [noise_mask, torch.ones((mb, mc, pad_latents, mh, mw), dtype=noise_mask.dtype, device=noise_mask.device)],
+                dim=2,
+            )
+
+        latent = {"samples": latent_image, "noise_mask": noise_mask}
+
+        # CONDITIONING PATH: slideshow -> IC-LoRA -> keep conditioning, discard latent.
+        cp, cn, _ = _unpack(_execute_comfy_node(
+            IC, positive=positive, negative=negative, vae=vae, latent=latent, image=slideshow,
+            frame_idx=0, strength=strength, latent_downscale_factor=downscale,
+            crop="center", use_tiled_encode=False, tile_size=256, tile_overlap=64,
+        ))
+
+        positive, negative = cp, cn
+
+        # LATENT PATH: one frozen keyframe per reference at frame 0 -> keep latent, drop cond.
+        for kf in keyframes:
+            _p, _n, latent = _unpack(_execute_comfy_node(
+                AG, positive=positive, negative=negative, vae=vae, latent=latent,
+                image=kf, frame_idx=0, strength=strength,
+            ))
+
+        return io.NodeOutput(positive, negative, latent)