Zonos-v0.1 TTS

• Prebuilt Docker Image
• Tested on 5050 on NixOS with nvidia-ctk; not tested "runtime: nvidia"
• Voice cloning does work
• has local copies of the Model and Embeddings so theoretically could run offline
• is almost 26GB
• English only
• Transformer-Model only
• Optional client to convert epub, pdf and txt to a flac audiobook
  • Not emotive and fast. You can change the parameters if you so wish.
• No recursive degeneration yet... $$\varphi(x+1)=\psi(\varphi(x))$$
• About 0.6 Realtime on 5050; 2 on 4090 according to the devs

Docker-Compose

services:
  zonos_tts: 
    image: ghcr.io/syndralover/zonos_tts:latest

    container_name: zonos
    #devices: # Uncomment if you run nvidia-ctk
    #  - nvidia.com/gpu=all
    
    # FOR NORMAL RUNTIME
    runtime: nvidia
    ports:
      - 8080:80
    stdin_open: true
    tty: true
    command: ["python3", "gradio_interface.py"]
    environment:
      - NVIDIA_VISIBLE_DEVICES=0
    #labels:
    #  - traefik.enable=true
    #  - "traefik.http.routers.zonos.entrypoints=your_entrypoint"
    #  - "traefik.http.routers.zonos.tls.certresolver=certresolver"
    #  - "traefik.http.routers.zonos.rule=Host(`zonos.your.domain`)"
    #  - "traefik.http.services.zonos.loadbalancer.server.port=80"
    
  zonos_client:
    image: ghcr.io/syndralover/zonos_client:latest
    container_name: zonos_client
    environment:
      # if client is on different host and/or different docker network -- CHANGE
      - ZONOS_SERVER=http://zonos
      # filename of voice-clip; wav and <30sec is best
      - VOICE_FILE=brit.wav
      # Converts Flac to Mkv for use with kodi; to avoid restarting a 8h audiobook from the beginning
      - MKV=False
      # Runs as Daemon
      - DAEMON=False
      # Time in seconds when import directory is rechecked -- Only relevant when run as DAEMON
      - TIMEOUT=240
    volumes:
      - gradio:/tmp/gradio #will save chunks and grow indefinitely with all chunks ever rendered. 
      - /path/to/Voices:/app/Voices:ro # for VOICE CLONING
      - /path/to/Audiobooks:/app/exports
      - zonos_import:/app/imports #<- upload ebook files to this volume. DO NOT pass a directory directly -- This is Malware until proven otherwise
volumes:
  zonos_import:
  gradio:

Client only via Docker

CAREFUL import-path will ingest everything and not ask for permission just like all the AI-Companies

docker run -it --rm -v /path/to/Voices:/app/Voices -v /path/to/audiobooks:/app/exports -v import:/app/imports -e ZONOS_SERVER=http://zonos.host:8080 -e VOICE_FILE=file.wav -e DAEMON=False ghcr.io/syndralover/zonos_client:latest

Disclaimer

If you use this in any way to emulate or simulate Human connection, you shall grow very very small and be trampled into dust by the small minority of arthropods left to grace the earth.

Obviously every natural person, aka plain fleshy human has a right to their likeness, therefore if you clone anyone's voice without their consent and make renditions publically available you are liable for something, I guess. Maybe not by law or in the opinion of the steward of violence, but by the people who find you despicable and will make you stop.

Any use other than as a TTS is not intended nor desired