svjack/YiChen_FramePack_lora_early

FramePack Dancing Image-to-Video Generation

This repository contains the necessary steps and scripts to generate videos using the Dancing image-to-video model. The model leverages LoRA (Low-Rank Adaptation) weights and pre-trained components to create high-quality anime-style videos based on textual prompts.

Prerequisites

Before proceeding, ensure that you have the following installed on your system:

• Ubuntu (or a compatible Linux distribution) • Python 3.x • pip (Python package manager) • Git • Git LFS (Git Large File Storage) • FFmpeg

Installation

1. Update and Install Dependencies

   sudo apt-get update && sudo apt-get install cbm git-lfs ffmpeg
   

2. Clone the Repository

   git clone https://huggingface.co/svjack/YiChen_FramePack_lora_early
   cd YiChen_FramePack_lora_early
   

3. Install Python Dependencies

   pip install torch torchvision
   pip install -r requirements.txt
   pip install ascii-magic matplotlib tensorboard huggingface_hub datasets
   pip install moviepy==1.0.3
   pip install sageattention==1.0.6
   

4. Download Model Weights

    git clone https://huggingface.co/lllyasviel/FramePackI2V_HY
    git clone https://huggingface.co/hunyuanvideo-community/HunyuanVideo
    git clone https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged
    git clone https://huggingface.co/Comfy-Org/sigclip_vision_384
   

Usage

To generate a video, use the fpack_generate_video.py script with the appropriate parameters. Below are examples of how to generate videos using the Dancing model.

1. Furina

• Source Image

python fpack_generate_video.py \
    --dit FramePackI2V_HY/diffusion_pytorch_model-00001-of-00003.safetensors \
    --vae HunyuanVideo/vae/diffusion_pytorch_model.safetensors \
    --text_encoder1 HunyuanVideo_repackaged/split_files/text_encoders/llava_llama3_fp16.safetensors \
    --text_encoder2 HunyuanVideo_repackaged/split_files/text_encoders/clip_l.safetensors \
    --image_encoder sigclip_vision_384/sigclip_vision_patch14_384.safetensors \
    --image_path fln.png \
    --prompt "In the style of Yi Chen Dancing White Background , The character's movements shift dynamically throughout the video, transitioning from poised stillness to lively dance steps. Her expressions evolve seamlessly—starting with focused determination, then flashing surprise as she executes a quick spin, before breaking into a joyful smile mid-leap. Her hands flow through choreographed positions, sometimes extending gracefully like unfolding wings, other times clapping rhythmically against her wrists. During a dramatic hip sway, her fingers fan open near her cheek, then sweep downward as her whole body dips into a playful crouch, the sequins on her costume catching the light with every motion." \
    --video_size 960 544 --video_seconds 3 --fps 30 --infer_steps 25 \
    --attn_mode sdpa --fp8_scaled \
    --vae_chunk_size 32 --vae_spatial_tile_sample_min_size 128 \
    --save_path save --output_type both \
    --seed 1234 --lora_multiplier 1.0 --lora_weight framepack_yichen_output/framepack-yichen-lora-000006.safetensors

• Without Lora

• With Lora

2. Roper

• Source Image

python fpack_generate_video.py \
    --dit FramePackI2V_HY/diffusion_pytorch_model-00001-of-00003.safetensors \
    --vae HunyuanVideo/vae/diffusion_pytorch_model.safetensors \
    --text_encoder1 HunyuanVideo_repackaged/split_files/text_encoders/llava_llama3_fp16.safetensors \
    --text_encoder2 HunyuanVideo_repackaged/split_files/text_encoders/clip_l.safetensors \
    --image_encoder sigclip_vision_384/sigclip_vision_patch14_384.safetensors \
    --image_path shengjiang.png \
    --prompt "In the style of Yi Chen Dancing White Background , The character's movements shift dynamically throughout the video, transitioning from poised stillness to lively dance steps. Her expressions evolve seamlessly—starting with focused determination, then flashing surprise as she executes a quick spin, before breaking into a joyful smile mid-leap. Her hands flow through choreographed positions, sometimes extending gracefully like unfolding wings, other times clapping rhythmically against her wrists. During a dramatic hip sway, her fingers fan open near her cheek, then sweep downward as her whole body dips into a playful crouch, the sequins on her costume catching the light with every motion." \
    --video_size 960 544 --video_seconds 3 --fps 30 --infer_steps 25 \
    --attn_mode sdpa --fp8_scaled \
    --vae_chunk_size 32 --vae_spatial_tile_sample_min_size 128 \
    --save_path save --output_type both \
    --seed 1234 --lora_multiplier 1.0 --lora_weight framepack_yichen_output/framepack-yichen-lora-000006.safetensors

• With Lora

3. Varesa

• Source Image

python fpack_generate_video.py \
    --dit FramePackI2V_HY/diffusion_pytorch_model-00001-of-00003.safetensors \
    --vae HunyuanVideo/vae/diffusion_pytorch_model.safetensors \
    --text_encoder1 HunyuanVideo_repackaged/split_files/text_encoders/llava_llama3_fp16.safetensors \
    --text_encoder2 HunyuanVideo_repackaged/split_files/text_encoders/clip_l.safetensors \
    --image_encoder sigclip_vision_384/sigclip_vision_patch14_384.safetensors \
    --image_path waliesha.jpg \
    --prompt "In the style of Yi Chen Dancing White Background , The dancer’s energy pulses in waves—one moment a statue, poised and precise, the next a whirl of motion as her feet flicker across the floor. Her face tells its own story: brows knit in concentration, then eyes widening mid-turn as if startled by her own speed, before dissolving into laughter as she springs upward, weightless. Her arms carve the air—now arcing like ribbons unfurling, now snapping sharp as a whip’s crack, palms meeting wrists in staccato beats. A roll of her hips sends her fingers fluttering near her temple, then cascading down as she folds into a teasing dip, the beads on her dress scattering light like sparks." \
    --video_size 960 544 --video_seconds 3 --fps 30 --infer_steps 25 \
    --attn_mode sdpa --fp8_scaled \
    --vae_chunk_size 32 --vae_spatial_tile_sample_min_size 128 \
    --save_path save --output_type both \
    --seed 1234 --lora_multiplier 1.0 --lora_weight framepack_yichen_output/framepack-yichen-lora-000006.safetensors

• With Lora

4. Scaramouche

• Source Image

python fpack_generate_video.py \
    --dit FramePackI2V_HY/diffusion_pytorch_model-00001-of-00003.safetensors \
    --vae HunyuanVideo/vae/diffusion_pytorch_model.safetensors \
    --text_encoder1 HunyuanVideo_repackaged/split_files/text_encoders/llava_llama3_fp16.safetensors \
    --text_encoder2 HunyuanVideo_repackaged/split_files/text_encoders/clip_l.safetensors \
    --image_encoder sigclip_vision_384/sigclip_vision_patch14_384.safetensors \
    --image_path shanbing.jpg \
    --prompt "In the style of Yi Chen Dancing White Background , The dancer’s energy pulses in waves—one moment a statue, poised and precise, the next a whirl of motion as her feet flicker across the floor. Her face tells its own story: brows knit in concentration, then eyes widening mid-turn as if startled by her own speed, before dissolving into laughter as she springs upward, weightless. Her arms carve the air—now arcing like ribbons unfurling, now snapping sharp as a whip’s crack, palms meeting wrists in staccato beats. A roll of her hips sends her fingers fluttering near her temple, then cascading down as she folds into a teasing dip, the beads on her dress scattering light like sparks." \
    --video_size 960 544 --video_seconds 3 --fps 30 --infer_steps 25 \
    --attn_mode sdpa --fp8_scaled \
    --vae_chunk_size 32 --vae_spatial_tile_sample_min_size 128 \
    --save_path save --output_type both \
    --seed 1234 --lora_multiplier 1.0 --lora_weight framepack_yichen_output/framepack-yichen-lora-000006.safetensors

• With Lora

Parameters

• --fp8: Enable FP8 precision (optional).
• --task: Specify the task (e.g., t2v-1.3B).
• --video_size: Set the resolution of the generated video (e.g., 1024 1024).
• --video_length: Define the length of the video in frames.
• --infer_steps: Number of inference steps.
• --save_path: Directory to save the generated video.
• --output_type: Output type (e.g., both for video and frames).
• --dit: Path to the diffusion model weights.
• --vae: Path to the VAE model weights.
• --t5: Path to the T5 model weights.
• --attn_mode: Attention mode (e.g., torch).
• --lora_weight: Path to the LoRA weights.
• --lora_multiplier: Multiplier for LoRA weights.
• --prompt: Textual prompt for video generation.

Output

The generated video and frames will be saved in the specified save_path directory.

Troubleshooting

• Ensure all dependencies are correctly installed. • Verify that the model weights are downloaded and placed in the correct locations. • Check for any missing Python packages and install them using pip.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

• Hugging Face for hosting the model weights. • Wan-AI for providing the pre-trained models. • DeepBeepMeep for contributing to the model weights.

Contact

For any questions or issues, please open an issue on the repository or contact the maintainer.

---