Wyoming OVOS TTS Bridge

This bridge exposes any OpenVoiceOS TTS plugin as a Wyoming protocol server. Use it with Home Assistant, Rhasspy, and other Wyoming-compatible voice pipelines.

                         ┌──────────────────────────────────────┐
  Wyoming client         │         wyoming-ovos-tts              │
  (Home Assistant,       │                                      │
   Rhasspy, etc.)        │  ┌────────────────────────────────┐  │
                         │  │   OVOSTTSEventHandler          │  │
  Synthesize ───────────►│  │                                │  │
  AudioStart ◄───────────│  │  asyncio.to_thread() ──────────►│  OVOSTTSFactory
  AudioChunk* ◄──────────│  │  TTS.synth(text)               │  └──> TTS plugin
  AudioStop  ◄───────────│  │                                │
                         │  └────────────────────────────────┘  │
  Describe ─────────────►│  Info(tts=[TtsProgram(...)])         │
  Info ◄─────────────────│                                      │
                         └──────────────────────────────────────┘

Features

• Non-streaming TTS: a single Synthesize event returns the complete audio.
• Streaming TTS (Wyoming v1.7+): SynthesizeStart/SynthesizeChunk/SynthesizeStop streams audio at each sentence boundary, using sentence-stream (the same segmenter used by wyoming-piper), for lower latency.
• Multi-language: the bridge advertises the plugin's available_languages in the Wyoming Info response.
• Thread-safe: the blocking tts.synth() call runs via asyncio.to_thread(), so the event loop stays responsive.
• Error reporting: failures come back as Wyoming Error events.
• Signal handling: the bridge shuts down cleanly on SIGINT/SIGTERM.

Installation

From PyPI

pip install wyoming-ovos-tts

You also need the OVOS TTS plugin you want to bridge, for example:

pip install ovos-tts-plugin-server
pip install ovos-tts-plugin-piper

From source

git clone https://github.com/OpenVoiceOS/wyoming-ovos-tts.git
cd wyoming-ovos-tts
pip install -e .

Configuration

The bridge reads plugin configuration from mycroft.conf, under tts.<plugin-name>:

{
  "lang": "en-US",
  "tts": {
    "ovos-tts-plugin-server": {
      "host": "https://pipertts.ziggyai.online"
    },
    "ovos-tts-plugin-piper": {
      "voice": "en_US-lessac-medium"
    },
    "ovos-tts-plugin-espeak": {
      "voice": "en+f3"
    }
  }
}

The language comes from tts.<plugin-name>.lang if set, otherwise from lang at the root level. Each plugin's config section must match the value passed to --plugin-name.

Usage

# TCP server using public OVOS TTS servers
wyoming-ovos-tts --uri tcp://0.0.0.0:7892 \
                 --plugin-name ovos-tts-plugin-server \
                 --debug

# With streaming disabled (requires wyoming info)
wyoming-ovos-tts --uri tcp://0.0.0.0:7892 \
                 --plugin-name ovos-tts-plugin-piper \
                 --no-streaming

# Unix socket
wyoming-ovos-tts --uri unix:///run/wyoming-tts.sock \
                 --plugin-name ovos-tts-plugin-server

# Over stdio (for subprocess usage)
wyoming-ovos-tts --plugin-name ovos-tts-plugin-server

CLI Reference

Wyoming Protocol

Non-streaming flow

Client → Describe
Server → Info(tts=[TtsProgram(supports_synthesize_streaming=True, ...)])

Client → Synthesize(text="Hello world", voice=VoiceSettings(...))
Server → AudioStart(rate=22050, width=2, channels=1)
       → AudioChunk (1024 samples of PCM)
       → AudioChunk ...
       → AudioStop

Streaming flow (v1.7+)

Client → SynthesizeStart
Client → SynthesizeChunk(text="Hello. ")
Server → AudioStart → AudioChunk+ → AudioStop   (sentence "Hello.")
Client → SynthesizeChunk(text="How are you? I'm ")
Server → AudioStart → AudioChunk+ → AudioStop   (sentence "How are you?")
Client → SynthesizeChunk(text="fine.")
Client → SynthesizeStop
Server → AudioStart → AudioChunk+ → AudioStop   (sentence "I'm fine.")
Server → SynthesizeStopped

When --no-streaming is not set, sentence-stream segments incoming text into complete sentences. It handles abbreviations (Dr.), decimals (3.14), ellipses, and non-Latin scripts correctly. Each sentence is synthesized and streamed as its own AudioStart→AudioChunk→AudioStop group as soon as it is complete. This lowers time-to-first-audio compared to the non-streaming path.

A trailing partial sentence is flushed when SynthesizeStop arrives, followed by a single SynthesizeStopped.

Supported Plugin Types

The bridge supports any OVOS TTS plugin that implements TTS from ovos_plugin_manager.templates.tts:

• ovos-tts-plugin-server: proxy to remote TTS servers
• ovos-tts-plugin-piper: local Piper TTS
• ovos-tts-plugin-espeak: eSpeak NG synthesizer
• ovos-tts-plugin-mimic: Mycroft Mimic
• ovos-tts-plugin-google-tx: Google Translate TTS
• ovos-tts-plugin-nos: NOS TTS
• ovos-tts-plugin-sam: Software Automatic Mouth
• ovos-tts-plugin-azure: Microsoft Azure Cognitive Services
• ovos-tts-plugin-ibm: IBM Watson TTS
• ovos-tts-plugin-amazon: Amazon Polly TTS

Documentation

Detailed docs live in docs/:

• Configuration
• Home Assistant
• Wyoming protocol

Credits

TigreGótico develops this project for OpenVoiceOS.

This project was funded through the NGI0 Commons Fund, a fund established by NLnet with financial support from the European Commission's Next Generation Internet programme, under the aegis of DG Communications Networks, Content and Technology under grant agreement No 101135429.