LTX 2.3 Multiple Subject Reference

--- license: apache-2.0 tags: - video-generation - multi-reference - LTX-2.3 base_model: - Lightricks/LTX-2.3 github: https://github.com/liconstudio/ComfyUI-Licon-MSR library_name: diffusers --- # Licon MSR V2 for LTX-2.3 ## What's New in V2 Compared with V1, **Licon MSR V2** introduces significant improvements in three key areas: ### 1. Improved Consistency - Better preservation of character identity, clothing, objects, and scene details - More consistent appearance across frames - Improved alignment between multiple reference images and the generated video - Reduced identity drift and reference attribute loss ### 2. Improved Stability - More reliable results across repeated sampling runs - Reduced visual artifacts, flickering, and temporal inconsistencies - More stable generation in complex multi-subject compositions - Improved handling of motion and interactions between subjects ### 3. Improved Scene Logic - Better understanding of spatial and action relationships described in prompts - More natural subject positioning and interaction - Improved temporal progression from the beginning to the end of a video - More coherent composition of characters, objects, and backgrounds ## Overview This model implements a novel approach to multi-reference video generation using **Multiple Subject Reference (MSR)**. Instead of introducing additional encoder branches or fusion modules, we transform multiple static reference images into a pseudo-video sequence that shares the same representation space as the target video. ## Usage This LoRA requires the **[ComfyUI-Licon-MSR](https://github.com/liconstudio/ComfyUI-Licon-MSR)** plugin for ComfyUI. A sample workflow is included in the model files for easy testing and experimentation. ## Key Features ### Multi-Reference Visual Memory - **Token-level reference preservation**: Multiple reference images are encoded as video latents, preserving fine-grained visual information at the token level instead of compressing them into a single embedding - **Native self-attention retrieval**: Target video tokens directly access reference tokens through the model's existing self-attention mechanism, with no additional architectural components required - **In-context conditioning**: References serve as visual memory within the main token sequence rather than as external conditioning inputs ### Flexible Reference Composit...

--- license: apache-2.0 tags: - video-generation - multi-reference - LTX-2.3 base_model: - Lightricks/LTX-2.3 github: https://github.com/liconstudio/ComfyUI-Licon-MSR library_name: diffusers --- ## Overview This model implements a novel approach to multi-reference video generation using **Multiple Subject Reference (MSR)**. Instead of introducing additional encoder branches or fusion modules, we transform multiple static reference images into a pseudo-video sequence that shares the same representation space as the target video. ## Usage This LoRA requires the **[ComfyUI-Licon-MSR](https://github.com/liconstudio/ComfyUI-Licon-MSR)** plugin for ComfyUI. A sample workflow is included in the model files for easy testing and experimentation. ## Key Features ### Multi-Reference Visual Memory - **Token-level reference preservation**: Multiple reference images are encoded as video latents, preserving fine-grained visual information at token level rather than compressing into a single embedding - **Native self-attention retrieval**: The target video tokens directly access reference tokens through the model's existing self-attention mechanism—no new architectural components needed - **In-context conditioning**: References serve as "visual memory" within the main token sequence, not as external conditioning inputs ### Flexible Reference Composition - **2 to 5 reference images**: Supports varying numbers of reference inputs with increasing complexity - **Complementary semantic roles**: Each reference image can carry different information: - Subject identity - Object/prop details - Scene/background - Local textures - Multiple viewpoints ## What It Can Do ### Identity Preservation Across References Generate videos where multiple reference identities are simultaneously preserved: - Multiple characters from different reference images - Character + object combinations - Object + scene compositions ### Relation-Based Composition Beyond mere identity preservation, the model can compose references based on textual relation descriptions: - Action interactions (handing, picking up, pushing) - Spatial relationships (left-right, foreground-background) - Temporal event structures (start → process → result) ### Cross-Reference Attribute Selection The model learns to selectively retrieve attributes from different references: - Face from reference A, clothing fr...

Source context: 7729 downloads · 260 likes · Library diffusers · Repo LiconStudio/LTX-2.3-Multiple-Subject-Reference