fantasy-talking

中文阅读

FantasyTalking: Realistic Talking Portrait Generation via Coherent Motion Synthesis

🔥 Latest News!!

• July 7, 2025: Released a user guide to solve the common issues reported by recent researchers and the community.
• July 5, 2025: Our work is accepted by ACM MM 2025!
• April 29, 2025: Our work is merged to ComfyUI-Wan ! Thank kijai for the update 👏!
• April 28, 2025: We released the inference code and model weights for audio conditions.

Quickstart

🛠️Installation

Clone the repo:

git clone https://github.com/Fantasy-AMAP/fantasy-talking.git
cd fantasy-talking

Install dependencies:

# Ensure torch >= 2.0.0
pip install -r requirements.txt
# Optional to install flash_attn to accelerate attention computation
pip install flash_attn

🧱Model Download

Models	Download Link	Notes
Wan2.1-I2V-14B-720P	🤗 Huggingface 🤖 ModelScope	Base model
Wav2Vec	🤗 Huggingface 🤖 ModelScope	Audio encoder
FantasyTalking model	🤗 Huggingface 🤖 ModelScope	Our audio condition weights

Download models using huggingface-cli:

pip install "huggingface_hub[cli]"
huggingface-cli download Wan-AI/Wan2.1-I2V-14B-720P --local-dir ./models/Wan2.1-I2V-14B-720P
huggingface-cli download facebook/wav2vec2-base-960h --local-dir ./models/wav2vec2-base-960h
huggingface-cli download acvlab/FantasyTalking fantasytalking_model.ckpt --local-dir ./models

Download models using modelscope-cli:

pip install modelscope
modelscope download Wan-AI/Wan2.1-I2V-14B-720P --local_dir ./models/Wan2.1-I2V-14B-720P
modelscope download AI-ModelScope/wav2vec2-base-960h --local_dir ./models/wav2vec2-base-960h
modelscope download amap_cvlab/FantasyTalking   fantasytalking_model.ckpt  --local_dir ./models

🔑 Inference

python infer.py  --image_path ./assets/images/woman.png --audio_path ./assets/audios/woman.wav

You can control the character's behavior through the prompt. The recommended range for prompt and audio cfg is [3-7]. You can increase the audio cfg to achieve more consistent lip-sync.

python infer.py  --image_path ./assets/images/woman.png --audio_path ./assets/audios/woman.wav --prompt "The person is speaking enthusiastically, with their hands continuously waving." --prompt_cfg_scale 5.0 --audio_cfg_scale 5.0

We present a detailed table here. The model is tested on a single A100.(512x512, 81 frames).

torch_dtype	num_persistent_param_in_dit	Speed	Required VRAM
torch.bfloat16	None (unlimited)	15.5s/it	40G
torch.bfloat16	7*10**9 (7B)	32.8s/it	20G
torch.bfloat16	0	42.6s/it	5G

Gradio Demo

We construct an online demo in Huggingface. For the local gradio demo, you can run:

pip install gradio spaces
python app.py

🧩 Community Works

We ❤️ contributions from the open-source community! If your work has improved FantasyTalking, please inform us. Or you can directly e-mail [email protected]. We are happy to reference your project for everyone's convenience.

🔗Citation

If you find this repository useful, please consider giving a star ⭐ and citation

@article{wang2025fantasytalking,
   title={FantasyTalking: Realistic Talking Portrait Generation via Coherent Motion Synthesis},
   author={Wang, Mengchao and Wang, Qiang and Jiang, Fan and Fan, Yaqi and Zhang, Yunpeng and Qi, Yonggang and Zhao, Kun and Xu, Mu},
   journal={arXiv preprint arXiv:2504.04842},
   year={2025}
 }

Acknowledgments

Thanks to Wan2.1, HunyuanVideo, and DiffSynth-Studio for open-sourcing their models and code, which provided valuable references and support for this project. Their contributions to the open-source community are truly appreciated.

Star History