ovos-docker-tts

Dockerfiles and Compose stacks that package each OpenVoiceOS text-to-speech plugin as a self-contained ovos-tts-server HTTP container. Each engine builds on a shared base image and exposes port 9666.

Install

This repository has no Python package and no install step. It only builds OCI images.

Build the base image first, then an engine image:

docker build -t smartgic/ovos-tts-server-base:alpha base/
docker build --build-arg TAG=alpha -t smartgic/ovos-tts-server-piper:alpha piper/

CUDA variants use Dockerfile.cuda and build on smartgic/ovos-tts-server-base-cuda. CUDA images exist for base, piper, and phoonnx.

Images publish to the Docker Hub smartgic/ namespace. The image labels still point back to this repository as the source.

Usage

Copy .env and set CONFIG_FOLDER, OVOS_USER, TZ, and VERSION for your setup, then start the engines you need with Compose:

docker compose up -d
docker compose -f docker-compose.cuda.yml up -d

docker-compose.yml runs the CPU engines (Piper, Mimic3, Mimic, Google Translate, SAM, Matxa, NOS, Kokoro, Phoonnx, Coqui, Bark) on host ports 8089-8098, each with its own cache and Gradio volumes mounted read-write and a read-only mount of ${CONFIG_FOLDER} into /home/${OVOS_USER}/.config/mycroft. docker-compose.cuda.yml runs the GPU-accelerated Piper engine and reserves an NVIDIA device.

Layout

• base/: shared layer: Debian, a Python virtual environment, the ovos user, espeak-ng, ffmpeg, and aiohttp. Every engine image builds FROM this.
• piper/, mimic3/, mimic/, google-tx/, sam/, matxa/, nos/, phoonnx/, kokoro/: one directory per TTS engine. Each installs ovos-tts-server plus the matching ovos-tts-plugin-* and sets ENTRYPOINT ovos-tts-server --engine <plugin>.

Related projects

• OpenVoiceOS/ovos-tts-server: the HTTP server each image runs.
• OpenVoiceOS/gh-automations: shared CI workflows used across OpenVoiceOS repositories.
• OpenVoiceOS community docs: broader OpenVoiceOS documentation referenced by the image labels.

License

This repository has no LICENSE file.