docs: split japanese and english
This commit is contained in:
Hiroaki Ogasawara
@@ -0,0 +1,61 @@
|
||||
# ComfyUI-FramePackWrapper_PlusOne
|
||||
|
||||
[ComfyUI-FramePackWrapper_PlusOne](https://github.com/tori29umai0123/ComfyUI-FramePackWrapper_PlusOne) は、[ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper)および[ComfyUI-FramePackWrapper_Plus](https://github.com/ShmuelRonen/ComfyUI-FramePackWrapper_Plus)から派生した、FramePackの1フレーム推論ノード(kisekaeichi対応)を含むフォークです。
|
||||
|
||||
本リポジトリは、 @tori29umai0123 氏の[依頼を受けて](https://x.com/tori29umai/status/1928692381735432320)公開用にフォークしました。
|
||||
|
||||
## 機能
|
||||
|
||||
- **1フレーム推論**: 基本的な1フレーム推論および、kisekaeichi方式に対応しています。技術的詳細は[musubi-tunerのドキュメント](https://github.com/kohya-ss/musubi-tuner/blob/main/docs/framepack_1f.md)を参照してください。
|
||||
- **F1サンプラー対応**: より高品質で時間的一貫性の高い動画生成のための改良されたF1方式を採用
|
||||
- **LoRA統合**: 適切な重み付けと融合オプションを備えたHunyuanVideo LoRAの完全サポート
|
||||
- **タイムスタンプ付きプロンプト**: 特定のタイムスタンプでプロンプトを変更できる動的動画の作成
|
||||
- **柔軟な入力オプション**: 参照画像と空の潜在空間の両方で完全なクリエイティブコントロールが可能
|
||||
- **解像度制御**: 最適な動画サイズのための自動バケット検出
|
||||
- **ブレンド制御**: タイムスタンプ間の異なるプロンプトのスムーズな遷移
|
||||
|
||||
### Not yet supported
|
||||
|
||||
- 1フレーム推論のうち、f-mc (one frame multi-control) 方式は未対応です。
|
||||
|
||||
## インストール
|
||||
|
||||
1. このリポジトリをComfyUIのcustom_nodesフォルダにクローンします:
|
||||
```bash
|
||||
cd ComfyUI/custom_nodes
|
||||
git clone https://github.com/xhiroga/ComfyUI-FramePackWrapper_PlusOne.git
|
||||
```
|
||||
|
||||
2. 必要な依存パッケージをインストールします:
|
||||
```bash
|
||||
pip install -r requirements.txt
|
||||
```
|
||||
|
||||
3. 必要なモデルファイルをダウンロードして、modelsフォルダに配置します:
|
||||
- FramePackI2V_HY: [HuggingFace Link](https://huggingface.co/lllyasviel/FramePackI2V_HY)
|
||||
- FramePack_F1_I2V_HY: [HuggingFace Link](https://huggingface.co/lllyasviel/FramePack_F1_I2V_HY_20250503)
|
||||
|
||||
## モデルファイル
|
||||
|
||||
### メインモデルオプション
|
||||
- [FramePackI2V_HY_fp8_e4m3fn.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/blob/main/FramePackI2V_HY_fp8_e4m3fn.safetensors) - Optimized fp8 version (smaller file size)
|
||||
- [FramePackI2V_HY_bf16.safetensors](https://huggingface.co/Kijai/HunyuanVideo_comfy/blob/main/FramePackI2V_HY_bf16.safetensors) - BF16 version (better quality)
|
||||
|
||||
### 必要なコンポーネント
|
||||
- **CLIP Vision**: [sigclip_vision_384](https://huggingface.co/Comfy-Org/sigclip_vision_384/tree/main)
|
||||
- **Text Encoder and VAE**: [HunyuanVideo_repackaged](https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/tree/main/split_files)
|
||||
|
||||
## Usage
|
||||
|
||||
[example_workflows](./example_workflows)を参照ください。
|
||||
|
||||
## License
|
||||
|
||||
[MIT License](LICENSE)
|
||||
|
||||
## クレジット
|
||||
|
||||
- [FramePack](https://github.com/lllyasviel/Fooocus-FramePack): @lllyasviel 氏によるオリジナルの実装です。
|
||||
- [ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper): @kijai 氏によるオリジナルの実装です。
|
||||
- [ComfyUI-FramePackWrapper_Plus](https://github.com/ShmuelRonen/ComfyUI-FramePackWrapper_Plus): @ShmuelRonen 氏によるF1対応のフォークです。
|
||||
- [ComfyUI-FramePackWrapper_PlusOne](https://github.com/tori29umai0123/ComfyUI-FramePackWrapper_PlusOne): @tori29umai0123 氏による1フレーム推論対応のフォークです。
|
||||
@@ -1,18 +1,14 @@
|
||||
# ComfyUI-FramePackWrapper_PlusOne
|
||||
|
||||
[ComfyUI-FramePackWrapper_PlusOne](https://github.com/tori29umai0123/ComfyUI-FramePackWrapper_PlusOne) is a fork by @tori29umai0123 derived from [ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper) and [ComfyUI-FramePackWrapper_Plus](https://github.com/ShmuelRonen/ComfyUI-FramePackWrapper_Plus/tree/main), which includes a 1-frame inference node for FramePack (compatible with kisekaeichi). This repository was forked for public release at the request of @tori29umai0123.
|
||||
**[日本語](./README.ja.md)**
|
||||
|
||||
<details>
|
||||
<summary>日本語</summary>
|
||||
[ComfyUI-FramePackWrapper_PlusOne](https://github.com/tori29umai0123/ComfyUI-FramePackWrapper_PlusOne) is a fork derived from [ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper) and [ComfyUI-FramePackWrapper_Plus](https://github.com/ShmuelRonen/ComfyUI-FramePackWrapper_Plus), containing FramePack's single-frame inference node (with kisekaeichi support).
|
||||
|
||||
[ComfyUI-FramePackWrapper_PlusOne](https://github.com/tori29umai0123/ComfyUI-FramePackWrapper_PlusOne) は、[ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper)および[ComfyUI-FramePackWrapper_Plus](https://github.com/ShmuelRonen/ComfyUI-FramePackWrapper_Plus/tree/main)から派生した、 @tori29umai0123 氏によるFramePackの1フレーム推論ノード(kisekaeichi対応)を含むフォーク
|
||||
です。本リポジトリは、 @tori29umai0123 氏の依頼を受けて公開用にフォークしました。
|
||||
</details>
|
||||
|
||||
An improved wrapper for the FramePack project that allows the creation of videos of any length based on reference images and LoRAs with F1 sampler.
|
||||
This repository was forked for public release at the request of @tori29umai0123 [as requested here](https://x.com/tori29umai/status/1928692381735432320).
|
||||
|
||||
## Features
|
||||
|
||||
- **1-Frame Inference**: Supports basic single frame inference and the kisekaeichi method. For technical details, please refer to the [musubi-tuner documentation](https://github.com/kohya-ss/musubi-tuner/blob/main/docs/framepack_1f.md).
|
||||
- **F1 Sampler Support**: Uses the improved F1 video generation method for higher quality and better temporal coherence
|
||||
- **LoRA Integration**: Full support for HunyuanVideo LoRAs with proper weight handling and fusion options
|
||||
- **Timestamped Prompts**: Create dynamic videos with changing prompts at specific timestamps
|
||||
@@ -20,12 +16,16 @@ An improved wrapper for the FramePack project that allows the creation of videos
|
||||
- **Resolution Control**: Automatic bucket finding for optimal video dimensions
|
||||
- **Blend Control**: Smooth transitions between different prompts at timestamps
|
||||
|
||||
### Not yet supported
|
||||
|
||||
- 1-Frame Inference: f-mc (one frame multi-control) is not supported yet.
|
||||
|
||||
## Installation
|
||||
|
||||
1. Clone this repository into your ComfyUI custom_nodes folder:
|
||||
```bash
|
||||
cd ComfyUI/custom_nodes
|
||||
git clone https://github.com/tori29umai0123/ComfyUI-FramePackWrapper_PlusOne.git
|
||||
git clone https://github.com/xhiroga/ComfyUI-FramePackWrapper_PlusOne.git
|
||||
```
|
||||
|
||||
2. Install the required dependencies:
|
||||
@@ -49,96 +49,7 @@ pip install -r requirements.txt
|
||||
|
||||
## Usage
|
||||
|
||||
### Basic Workflow
|
||||
|
||||
1. Load the HunyuanVideo model with your preferred settings
|
||||
2. (Optional) Add LoRAs with the FramePackLoraSelect node
|
||||
3. Prepare your input image or empty latent
|
||||
4. Set up CLIP Vision encoding for image embeddings
|
||||
5. Create timestamped prompts with FramePackTimestampedTextEncode
|
||||
6. Generate your video with FramePackSampler (F1)
|
||||
|
||||
### Example Workflow
|
||||
|
||||
|
||||
|
||||
### Timestamped Prompts
|
||||
|
||||
Use the following format for timestamped prompts:
|
||||
```
|
||||
[0s: A beautiful landscape, mountains in the background]
|
||||
[5s-10s: Camera pans to reveal a lake, reflections of clouds]
|
||||
[10s: A boat appears on the horizon, sailing slowly]
|
||||
```
|
||||
|
||||
- `[Xs: prompt]`: Starts at X seconds and continues until the next timestamp
|
||||
- `[Xs-Ys: prompt]`: Active from X seconds to Y seconds
|
||||
|
||||
### LoRA Usage
|
||||
|
||||
1. Place your HunyuanVideo LoRAs in the `ComfyUI/models/loras` folder
|
||||
2. Use the FramePackLoraSelect node to add them to your workflow
|
||||
3. Adjust strength as desired (typically 0.5-1.2)
|
||||
4. Set fuse_lora to false for flexibility or true for performance
|
||||
|
||||
## Node Reference
|
||||
|
||||
### FramePackSampler (F1)
|
||||
The main generation node using the F1 sampling technique.
|
||||
|
||||
**Inputs:**
|
||||
- `model`: The loaded FramePack model
|
||||
- `positive_timed_data`: Timestamped positive prompts
|
||||
- `negative`: Negative prompt conditioning
|
||||
- `start_latent`: Initial latent for generation
|
||||
- `start_image_embeds`: CLIP Vision embeddings for start image
|
||||
- `end_latent`: (Optional) End latent for transitions
|
||||
- `end_image_embeds`: (Optional) CLIP Vision embeddings for end image
|
||||
- `initial_samples`: (Optional) For video-to-video generation
|
||||
- Various sampling parameters (steps, CFG, guidance scale, etc.)
|
||||
|
||||
### FramePackTimestampedTextEncode
|
||||
Encodes text prompts with timestamps for timed conditioning.
|
||||
|
||||
**Inputs:**
|
||||
- `clip`: CLIP text model
|
||||
- `text`: Text prompt with timestamps
|
||||
- `negative_text`: Negative prompt
|
||||
- `total_second_length`: Video duration in seconds
|
||||
- `latent_window_size`: Processing window size
|
||||
- `prompt_blend_sections`: Number of sections to blend prompts
|
||||
|
||||
### FramePackLoraSelect
|
||||
Selects and configures LoRA models.
|
||||
|
||||
**Inputs:**
|
||||
- `lora`: LoRA model selection
|
||||
- `strength`: LoRA strength (0.0-2.0)
|
||||
- `fuse_lora`: Whether to fuse the LoRA weights into the base model
|
||||
- `prev_lora`: (Optional) For chaining multiple LoRAs
|
||||
|
||||
### LoadFramePackModel / DownloadAndLoadFramePackModel
|
||||
Loads the FramePack model with various precision options.
|
||||
|
||||
## Advanced Tips
|
||||
|
||||
1. **Resolution Control**: Use the FramePackFindNearestBucket node to optimize dimensions
|
||||
2. **Memory Management**: Adjust gpu_memory_preservation for large models
|
||||
3. **Blending Prompts**: Set prompt_blend_sections > 0 for smooth transitions
|
||||
4. **Multiple LoRAs**: Chain several LoRAs together for combined effects
|
||||
5. **Empty Latents**: Use an Empty Latent Image node when starting from scratch
|
||||
|
||||
## Troubleshooting
|
||||
|
||||
- **CUDA Out of Memory**: Reduce resolution, decrease latent_window_size, or increase gpu_memory_preservation
|
||||
- **LoRA Loading Issues**: Ensure LoRAs are in the correct format (safetensors)
|
||||
- **Video Artifacts**: Try increasing steps or adjusting CFG/guidance_scale
|
||||
|
||||
## Acknowledgements
|
||||
|
||||
- Based on the original [ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper) by kijai
|
||||
- Uses models from [FramePack](https://github.com/lllyasviel/Fooocus-FramePack) by lllyasviel
|
||||
- Special thanks to the ComfyUI and Stable Diffusion communities
|
||||
See [example_workflows](./example_workflows).
|
||||
|
||||
## License
|
||||
|
||||
@@ -146,12 +57,7 @@ Loads the FramePack model with various precision options.
|
||||
|
||||
## Credits
|
||||
|
||||
This project is an extension of the original [ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper) created by kijai. The original wrapper provided the foundation for working with FramePack models in ComfyUI.
|
||||
|
||||
ComfyUI-FramePackWrapper_Plus builds upon that foundation by adding support for:
|
||||
- F1 sampler for improved temporal coherence
|
||||
- LoRA integration for customized generation
|
||||
- Timestamped prompts for dynamic video creation
|
||||
- Additional workflow improvements and optimizations
|
||||
|
||||
Special thanks to kijai for the original implementation that made this extension possible.
|
||||
- [FramePack](https://github.com/lllyasviel/Fooocus-FramePack): @lllyasviel's original implementation.
|
||||
- [ComfyUI-FramePackWrapper](https://github.com/kijai/ComfyUI-FramePackWrapper): @kijai's original implementation.
|
||||
- [ComfyUI-FramePackWrapper_Plus](https://github.com/ShmuelRonen/ComfyUI-FramePackWrapper_Plus): @ShmuelRonen's F1-supported fork.
|
||||
- [ComfyUI-FramePackWrapper_PlusOne](https://github.com/tori29umai0123/ComfyUI-FramePackWrapper_PlusOne): @tori29umai0123's 1-frame inference-supported fork.
|
||||
|
||||
Reference in New Issue
Block a user