Skip to main content
This page documents the original speech endpoint. Start new integrations at the TTS quickstart: the current endpoint lives at /v1/audio/speech, uses model mira-tts, streams, and takes the same sk_live_ key as the rest of the API.

​
What /tts/v1 does now

The URL keeps working. POST /tts/v1/audio/speech is an alias for POST /v1/audio/speech — the same handler, the same audio — so a caller that only changes its key needs no other change:
  • Auth is your sk_live_ API key, the one you mint in the console under Developers. The old mira_ speech keys are retired; if you are still holding one, get an sk_live_ key from the quickstart.
  • model is mira-tts, and mira-tts-v51 is accepted as an alias, so an existing body keeps working unchanged.
  • Everything else follows the current endpoint — voices, pcm streaming, the 2,000-character ceiling, the response headers, the price, the wallet it debits, and the request log. One contract, documented in one place.
  • The options below that were unique to this endpoint — speech_ready, instructions, lexicon and numbers — are no longer applied. They are ignored, not rejected.
The rest of this page describes the original endpoint as it behaved, for callers reading old code.

​
The original endpoint

The speech endpoint is the voice from our agents, on its own. Send text, get a WAV back. It takes the same request shape as the OpenAI audio API, so any OpenAI SDK works by swapping base_url and the key. The engine speaks Hindi, English and the Hinglish in between, plus Gujarati in beta. Five more languages — Telugu, Tamil, Marathi, Kannada and Bengali — arrive over the next two months: see the language roadmap.

​
Calling the alias

The old body, the new key:
The response body is the audio: 48 kHz mono PCM16, wav by default. Ask for "response_format": "pcm" and it streams, which is what you want in a live pipeline — see the TTS quickstart.

​
Request

Read as history. On the alias, model, input, voice and response_format behave as the TTS quickstart documents — input goes up to 2,000 characters and pcm streams — and the four fields below that are ours alone are ignored.
model is a stable public name, not the build number. It always points at the current production voice, which we upgrade underneath you — so a clip you generate today may sound better than one from last month without your code changing.

​
The rewrite pass

The rewrite pass is not part of the current endpoint. speech_ready, instructions and lexicon are ignored and no X-Mira-Speech-Ready header is returned. Send text you are happy to hear read aloud.
Raw text is rarely speech-ready. 97% should be read as “ninety-seven percent”, a product name should not be transliterated, and a bare URL should not be spelled out character by character. By default we run your text through a rewrite that fixes exactly that, preserving meaning, before it reaches the engine. Every response carries a header naming what happened: fallback is a degradation, never an error: you still get audio and still get a 200. If you are debugging pronunciation and the header says fallback, the rewrite is not what shaped that clip.
Skip the rewrite

​
Listing models

The list names one model — the current one:
mira-tts-v51 is still accepted in a request body; it is no longer listed.

​
Errors

Errors use the same envelope as the rest of the API — see Errors.
404 Not Found
The alias returns the current endpoint’s codes instead — invalid_request, unauthorized, insufficient_balance, at_capacity, rate_limited, model_not_found, upstream_error. See When something fails.

​
Using an OpenAI SDK

The shape matches, so point the client at us and keep your code:
lexicon, numbers, instructions and speech_ready are ours, not OpenAI’s. SDKs that validate their request body may reject them — send those with a plain HTTP client, or use extra_body where your SDK supports it.