Files
2024-06-29 18:50:22 +08:00

52 lines
1.5 KiB
Markdown

---
tags:
- AnimateDiff
- Animation
---
# Custom CFG 🎭🅐🅓
## Documentation
- Class name: `ADE_CustomCFG`
- Category: `Animate Diff 🎭🅐🅓/sample settings`
- Output node: `False`
This node is designed for creating custom configuration settings for animation diffusion processes, allowing users to specify multiple values for configuration settings that influence the generation process.
## Input types
### Required
- **`cfg_multival`**
- Specifies the multiple values for configuration settings, enabling fine-tuned control over the animation diffusion process.
- Comfy dtype: `MULTIVAL`
- Python dtype: `Union[float, torch.Tensor]`
## Output types
- **`custom_cfg`**
- Comfy dtype: `CUSTOM_CFG`
- Outputs a custom configuration group, encapsulating the specified configuration settings for use in animation diffusion.
- Python dtype: `CustomCFGKeyframeGroup`
## Usage tips
- Infra type: `CPU`
- Common nodes: unknown
## Source code
```python
class CustomCFGNode:
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"cfg_multival": ("MULTIVAL",),
}
}
RETURN_TYPES = ("CUSTOM_CFG",)
CATEGORY = "Animate Diff 🎭🅐🅓/sample settings"
FUNCTION = "create_custom_cfg"
def create_custom_cfg(self, cfg_multival: Union[float, Tensor]):
keyframe = CustomCFGKeyframe(cfg_multival=cfg_multival)
cfg_custom = CustomCFGKeyframeGroup()
cfg_custom.add(keyframe)
return (cfg_custom,)
```