video-dubbing-svc/app/run.sh

#!/bin/bash

mkdir -p /data/output /data/ingest /data/models /data/temp /data/cache
rm -rf /data/temp/*

# For each video file in the ingest directory
for f in /data/ingest/*; do
    # Move file to temp directory
    mv "$f" /data/temp/
    # Get file name without extension
    filename="$(basename "${f%.*}")"
    # Extract audio from video file
    ffmpeg -loglevel warning -y -i /data/temp/"${f##*/}" -f wav /data/temp/"$filename".wav
    # Split voice and music from audio file
    python3 /vocal-remover/inference.py -i /data/temp/"$filename".wav -o /data/temp/ --tta --pretrained_model /vocal-remover/models/baseline.pth
    # Clone the voice
    svc infer -o /data/temp/"$filename"_"$SPEAKER".wav -m "$MODEL_PATH" -c "$MODEL_CONFIG_PATH" -s "$SPEAKER" -na -t "$TRANSPOSE" /data/temp/"$filename"_Vocals.wav
    # Combine voice and music into one mp3 file
    ffmpeg -loglevel warning -y -i /data/temp/"$filename"_"$SPEAKER".wav -i /data/temp/"$filename"_Instruments.wav \
        -filter_complex "[0:a]volume=$VOCALS_VOLUME[a0];[a0][1:a]amix=inputs=2:duration=longest" /data/temp/"$filename"_"$SPEAKER"_combined.wav
    # Combine audio and video into one file
    ffmpeg -loglevel warning -y -i /data/temp/"${f##*/}" -i /data/temp/"$filename"_"$SPEAKER"_combined.wav -c:v copy -map 0:v:0 -map 1:a:0 /data/output/"$filename"_"$SPEAKER".mp4
done