gopipertts

A small HTTP API wrapper for piper's texttospeech

Usage

The provided docker image embeds piper for easier use. Just run:

docker run \
    -v ./voices:/voices \ # This is where the voices will be saved to. Mounted for persistence
    -p 8080:8080 \
    nbr23/gopipertts

You can then navigate to http://localhost:8080/ to access the webui and start speaking text.

API Reference

List voices

GET /api/voices will return a json list of voices available for download and usage

Process text into speech

/api/tts will convert the text passed into an audio file. The output format depends on the outputFormat parameter (wav by default, mp3 if specified). This endpoint accepts POST and GET requests.

POST requests expect a json body like the following:

{
    "text": "Hello World",
    "speed": 1.0,
    "voice": "en_US-amy-low",
    "speaker": "",              // only available for select voices
    "outputFormat": "wav"       // also accepts "mp3" (requires ffmpeg)
}

GET requests expect the parameters text and optionally speed, voice, speaker and outputFormat to be passed as url query parameters.

Some usage examples:

curl -X POST -H "Content-Type: application/json" -d '{"text": "happy text to speaching!"}' 'http://localhost:8080/api/tts' | mpv -

curl -X POST -H "Content-Type: application/json" -d '{"text": "happy text to speaching!", "voice":"en_US-amy-low", "speed": 1.2}' 'http://localhost:8080/api/tts' | mpv -

curl -X POST -H "Content-Type: application/json" -d '{"text": "happy text to speaching!"}' 'http://localhost:8080/api/tts?speed=1.2&voice=en_US-amy-low' | mpv -

curl 'http://localhost:8080/api/tts?speed=1.1&voice=en_US-amy-low&text=Happy%20text%20to%20speeching' | mpv -

curl -X POST -H "Content-Type: application/json" -d '{"text": "happy text to speaching!", "outputFormat": "mp3"}' 'http://localhost:8080/api/tts' | mpv -

Leverages piper for TTS and voices from rhasspy/piper-voices

Environment Variables

Variable	Default	Description
VOICES_PATH	/voices	Path to the voices directory
VOICES_JSON_PATH	/app/voices.json	Path to the voices metadata JSON file
STREAM_EXPIRATION_MINUTES	15	How long to cache audio streams
PRELOAD_VOICES		Comma-separated list of voices to preload on startup
LOG_INPUT		When set, prints TTS input text to stdout before synthesis
PORT	8080	HTTP port to listen on