This repository contains the demo for the audio-to-video synchronisation network (SyncNet). This network can be used for audio-visual synchronisation tasks including:
- Removing temporal lags between the audio and visual streams in a video;
- Determining who is speaking amongst multiple faces in a video.
Please cite the paper below if you make use of the software.
GPU (CUDA):
conda env create -f environment.yml
CPU only:
conda env create -f environment-cpu.yml
The code automatically detects and uses a CUDA GPU if available, and falls back to CPU otherwise.
Download the pretrained model:
sh download_model.sh
SyncNet demo:
python demo_syncnet.py --videofile data/example.avi --tmp_dir /path/to/temp/directory
Check that this script returns approximately the following values (minor differences are expected depending on your platform and package versions):
AV offset: 3
Min dist: 5.353
Confidence: 10.021
Run the three stages — face detection and tracking, sync offset estimation, and visualisation:
python run_pipeline.py --videofile /path/to/video.mp4 --reference name_of_video --data_dir /path/to/output
python run_syncnet.py --videofile /path/to/video.mp4 --reference name_of_video --data_dir /path/to/output
python run_visualise.py --videofile /path/to/video.mp4 --reference name_of_video --data_dir /path/to/output
Outputs:
$DATA_DIR/pycrop/$REFERENCE/*.avi - cropped face tracks
$DATA_DIR/pyavi/$REFERENCE/video_out.avi - output video (as shown below)
@InProceedings{Chung16a,
author = "Chung, J.~S. and Zisserman, A.",
title = "Out of time: automated lip sync in the wild",
booktitle = "Workshop on Multi-view Lip-reading, ACCV",
year = "2016",
}