Documentation

Speak

Text to speech in one HTTP call: 15 voices, 31 languages, up to 10,000 characters a request. Your Mosaic API key works as is. On Pro and Teams (pricing).

Make speech

curl -X POST "https://speak.mosaic.site/v1/speak?voice=eve" \
  -H "X-API-Key: $KEY" \
  -d '{"text": "Hello! I can talk now."}' \
  -o speech.wav

Speak lives at https://speak.mosaic.site. Send the key as X-API-Key or as Authorization: Bearer.

Parameters

In the query string or the JSON body; the query string wins.

Param	Default	Values
text	required	Up to 10,000 characters
voice	`eve`	See Voices below
lang	`en`	31 ISO codes, or `na` for unknown or mixed
speed	`1.05`	`0.7` (slow) to `2.0` (fast)
quality	`balanced`	`fast`, `balanced`, `high`
format	`wav`	`wav`, `pcm`, `mulaw`
sample_rate	by format	`8000`, `16000`, `44100`; `8000` for `mulaw`, `44100` otherwise
stream	`false`	`true` sends audio in chunks as each is ready

Voices

English — eve, alba, anna, caro_davy, paul. Fixed speed: any speed other than 1.05 returns 400 unsupported_for_voice.
All 31 languages — F1 to F5 and M1 to M5. Adjustable speed, and expression tags such as <laugh> and <sigh> inline in the text.
Without a voice, you get eve, or F1 when the request asks for another language or speed. A voice you name is never swapped.
GET /v1/voices lists the voices with samples; GET /v1/languages lists the language codes.

Phone audio

For phone greetings and voice agents, ask for 8 kHz μ-law with format=mulaw&sample_rate=8000:

curl -X POST "https://speak.mosaic.site/v1/speak?voice=eve&format=mulaw&sample_rate=8000" \
  -H "X-API-Key: $KEY" \
  -d '{"text": "Thanks for calling. How can I help?"}' \
  -o greeting.ulaw

For a greeting saved to a file, leave stream off: you get a proper error status and the cache. Use stream=true only to play audio as it arrives. A stream can end early without an error, so treat an unexpectedly short one as a failure.

Limits and usage

10,000 characters a request and 60 requests a minute per account.
GET /v1/usage returns {"requests", "characters"} for your account.

Errors

Errors return {"error": {"code": "...", "message": "..."}}.

Code	What it means
missing_key, invalid_key	401: no key, or the key isn't recognised
text_too_long	Over 10,000 characters
unsupported_for_voice	400: an English voice with another language or speed; use an F or M voice
rate_limited	429: over 60 requests a minute; back off
model_loading	503: the service just restarted; retry in about 30 seconds
platform_unavailable	503: Mosaic is briefly unavailable; retry shortly

Endpoints

On https://speak.mosaic.site:

Endpoint	Method	Description
/v1/speak	POST	Text in, audio out
/v1/voices	GET	Voices, with sample URLs
/v1/languages	GET	Language codes
/v1/usage	GET	Requests and characters for your account

← PreviousEmail Next →Git