RVC v2 + crepe

RVC v2 with Crepe, for AI music covers.

How to run

Using Docker

docker run \
    --name="rvc-v2-crepe" \
    -v ./logs:/app/logs \
    -v ./weights:/app/weights \
    -v ./inputs:/app/inputs \
    -v ./outputs:/app/audio-outputs \
    -v ./temp-outputs:/app/TEMP/gradio \
    -p "7865:7865" \
    --gpus="all" \
    ilshidur/rvc-2.0-crepe

Using Docker compose

# docker-compose.yml
version: '3.3'

services:
  rvc-2-0-crepe:
    image: ilshidur/rvc-2.0-crepe
    container_name: rvc-2-0-crepe
    volumes:
      - ./logs:/app/logs
      - ./weights:/app/weights
      - ./inputs:/app/inputs
      - ./outputs:/app/audio-outputs
      - ./temp-outputs:/app/TEMP/gradio
    ports:
      - "7865:7865"
    deploy: # --gpus all
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

Then run :

docker compose up --build

Exposed volumes

• /app/logs (Optional) : .index files. Recommended to follow the following pattern : logs/<modelname>/<arbitrary name>.index.
• /app/weights : the .pth files containing the voice models.
• /app/inputs : the input .wav files to infer.
• /app/audio-outputs : the final audio outputs.
• /app/TEMP/gradio (Optional) : temporary audio outputs.

UI

-> http://localhost:7865

As this is a Gradio app, you can interact with it using the WebSocket API to automate voice cloning.

Querying using the Gradio API

Check out the example folder to have an insight of a Node.js client infering an audio file using a pre-existing model. Please note that this example is just a code sample, not a fully working project.