LipNet-JP

Directory

src/

.py files here.

notebook/

.ipynb files here.

Prepare video

youtube-dl -F https://www.youtube.com/watch?v={youtube_id}
youtube-dl -f 22 -o 'data/raw_input/%(id)s.%(ext)s' https://www.youtube.com/watch?v={youtube_id}
ffmpeg -i {youtube_id}.mp4 -c:v copy {youtube_id}.avi

これでdockerを立ち上げて，中に入る．psでid確認してcpなどする．

docker run -it --rm algebr/openface:latest

docker cp {youtube_id}.avi {docker_id}:/home/openface-build

Docker の中

build/bin/FeatureExtraction -simsize 200 -f {youtube_id}.avi
build/bin/FeatureExtraction -simsize 200 -fdir processed/{youtube_id}_aligned/ -out_dir processed2

Dockerの外

docker cp {docker_id}:/home/openface-build/processed data/.
docker cp {docker_id}:/home/openface-build/processed2 data/.

python src/crop_lip.py {youtube_id}

Prepare audio

ffmpeg -y -i {youtube_id}.mp4 -c:a flac audio2.flac

sample2を適宜ファイル名に読み替える