Skip to main content
Version: v5.4.0

KittenTTS-FastAPI

Run KittenTTS-FastAPI locally and connect it to OpenReader using the Custom OpenAI-Like provider. Lightweight and CPU-friendly.

Run KittenTTS​

docker run -it --rm \
  --name kittentts-fastapi \
  -e KITTEN_MODEL_REPO_ID="KittenML/kitten-tts-nano-0.8-fp32" \
  -p 8005:8005 \
  ghcr.io/richardr1126/kittentts-fastapi-cpu

Connect to OpenReader​

In Settings → Admin → Providers:

  1. Add a provider with type Custom OpenAI-like.
  2. Set the base URL for your deployment topology (Docker-to-host: http://host.docker.internal:8005/v1; native same-host: http://127.0.0.1:8005/v1).
  3. Leave API key blank unless required by your deployment.
  4. Set default model to kitten-tts (or your backend model id).

Or seed it on first boot (see Seed a provider from the environment):

API_BASE=http://host.docker.internal:8005/v1

Use host.docker.internal only when OpenReader/its embedded worker run in Docker and KittenTTS runs on that Docker host. In Compose, use the provider service name on the shared network. A remote worker needs a public/private-network URL it can reach. See the provider address table.

Users choose the provider, model and voice in Settings → Model & Voice.

References​