Phonon TTS for Unraid

An Unraid Community Applications template for ptts-openai-server, the official speech server for Phonon, Gradium's open-weight on-device text-to-speech models. It serves an OpenAI-compatible API (POST /v1/audio/speech) and the Wyoming protocol, so it plugs into Open WebUI, SillyTavern, Home Assistant and anything else with an "OpenAI TTS" setting — fully local, CPU-only, and fast enough to speak faster than real time even on a Raspberry Pi.

Report template or Unraid problems in this repository's issues, and problems with the server itself to the xn-ptts project.

Install

Search for Phonon TTS in the Apps tab. The defaults work out of the box:

• API port 8880 — point any OpenAI-TTS client at http://<tower>:8880/v1
• Language en — text normalization and the catalog voices; must match the model's language
• Model and voices cache — under about 1 GB, downloaded on first start and kept across restarts and updates

The first start downloads the default model and voices and logs little until they load — watch the log. Later starts go straight to serving.

The server has no authentication on any port. Run it on a trusted network; never expose it to the internet.

Home Assistant

Set Wyoming address to 0.0.0.0:10200 and map the Wyoming port too (both under Advanced). Then in Home Assistant add the Wyoming Protocol integration with this server's address and port 10200: Phonon appears as a text-to-speech engine you can pick for your voice assistant. It streams — speech starts at the first sentence of a reply, not the last.

Models and voices

• Default: leave Model empty — the server ships a built-in default model.
• Other models: hf://owner/name (add @commit_sha to pin the version). The Phonon 40M models are gated: accept their terms on Hugging Face first, then set an HF token.
• Voices: the public catalog (gradium/phonon-voices) is cached for your language; GET /v1/audio/voices lists what the running server speaks. OpenAI voice names and Voice aliases make clients with fixed voice lists (they send alloy, nova, …) work unchanged.
• Ignore speed is on by default: OpenAI clients always send a speed parameter; the server accepts it and speaks at the normal rate.

Updates

The template follows latest, which Gradium moves with each release; Unraid shows the update on the Docker tab. Pick a pinned tag (such as 0.5.0) under Branch at install to stay on one release.

Credits and licence

Phonon and the xn-ptts server are by Gradium and contributors, Apache-2.0. The models themselves carry their own licences on Hugging Face (the 31M models are Apache-2.0; the 40M models are gated — check their model cards). The icon in this repository is an original work by the template maintainer, MIT. The template files are MIT licensed; see LICENSE.