CLI

Download weight layers into local storage and register the model, installing its backend engine if needed.

• --force

  Pull even if already installed (resumes existing blobs)

• --skip-deps

  Don't install the backend engine

Examples

wavhost pull chatterbox-turbo
wavhost pull chatterbox-multilingual
wavhost pull qwen-0.6-customvoice
wavhost pull kokoro
wavhost pull qwen-0.6-base --force
wavhost pull chatterbox-turbo --skip-deps

Generate speech from text using a local model.

• -o, --output

  Output WAV path (default output.wav)

• --voice

  Saved voice, reference audio path, or named speaker (Qwen: Ryan, Aiden, …; Kokoro: af_heart, bm_george, …)

• --language, -l

  Multilingual ISO (en, fr, zh, …), Qwen names (English, …), or Kokoro ISO / voice prefix

• --device

  cuda, cpu, or mps

Examples

wavhost run chatterbox-turbo "Welcome to Wavhost" -o welcome.wav
wavhost run chatterbox-turbo "Hello" --voice my-voice -o hello.wav
wavhost run chatterbox-turbo "Hello" --voice /path/to/audio.wav -o hello.wav
wavhost run qwen-0.6-customvoice "Hello" --voice Ryan -o qwen.wav
wavhost run qwen-0.6-base "Hello" --voice /path/to/audio.wav -o clone.wav
wavhost run chatterbox-multilingual "Bonjour" --language fr -o fr.wav
wavhost run kokoro "Hello from Kokoro" --voice af_heart -o hello.wav

Create a saved voice from reference audio.

• --ref

  Required. Path to reference audio

• --desc

  Optional description

Example

Start the local API server.

• --hostdefault localhost

  Bind address

• --portdefault 11435

  Bind port

• --reloadoff

  Auto-reload for development

Examples

wavhost serve
wavhost serve --host 0.0.0.0 --port 8000 --reload