Update SKILL.md with accurate color correction documentation

- Replace hardcoded warning with explanation of why automatic detection isn't possible
- Document --invert_colors for luminosity inversion
- Document --swap_bgr for BGR<->RGB channel swapping
- Explain that VAE output format is model-specific
parent 539caee2
...@@ -535,13 +535,22 @@ Auto mode detects NSFW content automatically. For explicit requests: ...@@ -535,13 +535,22 @@ Auto mode detects NSFW content automatically. For explicit requests:
python3 videogen --model <model> --prompt "<prompt>" --no_filter python3 videogen --model <model> --prompt "<prompt>" --no_filter
``` ```
### 5. Color Inversion ### 5. Color Correction
Some models (particularly certain LoRA adapters) may produce inverted colors. Fix this with: Some models may produce videos with incorrect colors. Two manual fixes are available:
**Color inversion (luminosity)** - Dark areas appear light, light areas appear dark:
```bash ```bash
python3 videogen --model wan2_2_i2v_general_nsfw_lora --image input.png --prompt "..." --invert_colors python3 videogen --model wan2_2_i2v_general_nsfw_lora --image input.png --prompt "..." --invert_colors
``` ```
**BGR/RGB channel swap** - Colors have wrong tint (red appears blue, etc.):
```bash
python3 videogen --model wan2_2_i2v_general_nsfw_lora --image input.png --prompt "..." --swap_bgr
```
**Note:** There is no standard way to automatically detect RGB vs BGR output from diffusion models. The VAE output format is model-specific and not declared in metadata. Users must visually inspect output and apply corrections as needed.
### 6. Reproducibility ### 6. Reproducibility
Always capture and store the seed for reproducibility: Always capture and store the seed for reproducibility:
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment