Update README.md
Browse files
README.md
CHANGED
|
@@ -10,10 +10,6 @@ github: https://github.com/liconstudio/ComfyUI-Licon-MSR
|
|
| 10 |
library_name: diffusers
|
| 11 |
---
|
| 12 |
|
| 13 |
-
# Multiple Subject Reference IC-LoRA (Test Version)
|
| 14 |
-
|
| 15 |
-
> ⚠️ **This is a test version released for feedback collection to guide future optimization.**
|
| 16 |
-
|
| 17 |
## Overview
|
| 18 |
|
| 19 |
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.
|
|
@@ -64,23 +60,23 @@ The model learns to selectively retrieve attributes from different references:
|
|
| 64 |
- Object identity from one reference, pose/position from another
|
| 65 |
- Background elements from scene references
|
| 66 |
|
| 67 |
-
##
|
| 68 |
-
|
| 69 |
-
- **High-motion limb distortion**: Significant degradation in limb quality during fast or complex motion sequences
|
| 70 |
-
- **Slight object consistency loss**: Minor identity drift for objects throughout the video duration
|
| 71 |
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
|
| 74 |
## Results Showcase
|
| 75 |
|
| 76 |
-
###
|
| 77 |
|
| 78 |
-
|
| 79 |
-
|:---:|:---:|:---:|
|
| 80 |
-
| <img src="validation/1_1.png" width="120"> <img src="validation/1_2.png" width="120"> | [▶ Play](validation/1_video.mp4) | [▶ Play](validation/1_video_seed.mp4) |
|
| 81 |
|
| 82 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 83 |
|
| 84 |
-
|
| 85 |
-
|:---:|:---:|:---:|
|
| 86 |
-
| <img src="validation/2_1.png" width="80"> <img src="validation/2_2.png" width="80"> <img src="validation/2_3.png" width="80"> <img src="validation/2_4.png" width="80"> | [▶ Play](validation/2_video.mp4) | [▶ Play](validation/2_video_seed.mp4) |
|
|
|
|
| 10 |
library_name: diffusers
|
| 11 |
---
|
| 12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
## Overview
|
| 14 |
|
| 15 |
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.
|
|
|
|
| 60 |
- Object identity from one reference, pose/position from another
|
| 61 |
- Background elements from scene references
|
| 62 |
|
| 63 |
+
## Usage Tips (V1 Version)
|
|
|
|
|
|
|
|
|
|
| 64 |
|
| 65 |
+
- **Prompt description**: Requires concise but accurate description of reference images. Over-description or under-description both lead to consistency degradation
|
| 66 |
+
- **High-motion scenes**: 50fps recommended to ensure smooth motion coherence
|
| 67 |
+
- **Generation reliability**: Typically requires 2-3 sampling runs to achieve accurate results
|
| 68 |
|
| 69 |
|
| 70 |
## Results Showcase
|
| 71 |
|
| 72 |
+
### V1 Version
|
| 73 |
|
| 74 |
+
Character references + scene background composition:
|
|
|
|
|
|
|
| 75 |
|
| 76 |
+
| Reference Images | Generated Video |
|
| 77 |
+
|:---:|:---:|
|
| 78 |
+
| <img src="validition_v1/01/1.jpg" width="80"> <img src="validition_v1/01/2.jpg" width="80"> <img src="validition_v1/01/bg.png" width="80"> | <video src="validition_v1/01/video.mp4" controls width="400"></video> |
|
| 79 |
+
| <img src="validition_v1/07/1.jpg" width="80"> <img src="validition_v1/07/2.jpg" width="80"> <img src="validition_v1/07/bg.png" width="80"> | <video src="validition_v1/07/video.mp4" controls width="400"></video> |
|
| 80 |
+
| <img src="validition_v1/05/1.png" width="70"> <img src="validition_v1/05/2.png" width="70"> <img src="validition_v1/05/5.png" width="70"> <img src="validition_v1/05/bg.png" width="70"> | <video src="validition_v1/05/video.mp4" controls width="400"></video> |
|
| 81 |
|
| 82 |
+
---
|
|
|
|
|
|