Edit model card
YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/model-cards#model-card-metadata)

One-shot Talking Face Generation from Single-speaker Audio-Visual Correlation Learning (AAAI 2022)

Paper

Requirements

  • Python >= 3.6 , Pytorch >= 1.8 and ffmpeg
  • Set up OpenFace
    • We use the OpenFace tools to extract the initial pose of the reference image
    • Make sure you have installed this tool, and set the OPENFACE_POSE_EXTRACTOR_PATH in config.py. For example, it should be the absolute path of the "FeatureExtraction.exe" for Windows.
  • Other requirements are listed in the 'requirements.txt'

Pretrained Checkpoint

Please download the pretrained checkpoint from google-drive and unzip it to the directory (/checkpoints). Or manually modify the settings of GENERATOR_CKPT and AUDIO2POSE_CKPT in the config.py.

Extract phoneme

We employ the CMU phoneset to represent phonemes, the extra 'SIL' means silence. All the phonesets can be seen in 'phindex.json'.

We have extracted the phonemes for the audios in the 'sample/audio' directory. For other audios, you can extract the phonemes by other ASR tools and then map them to the CMU phoneset. Or email to [email protected] for help.

Generate Demo Results

python test_script.py --img_path xxx.jpg --audio_path xxx.wav --phoneme_path xxx.json --save_dir "YOUR_DIR"

Note that the input images must keep the same height and width and the face should be appropriately cropped as in samples/imgs. You can also preprocess your images with image_preprocess.py.

License and Citation

@InProceedings{wang2021one,
author = Suzhen Wang, Lincheng Li, Yu Ding, Xin Yu
title = {One-shot Talking Face Generation from Single-speaker Audio-Visual Correlation Learning},
booktitle = {AAAI 2022},
year = {2022},
}

Acknowledgement

This codebase is based on First Order Motion Model and imaginaire, thanks for their contributions.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference API
Unable to determine this model's library. Check the docs .