Text to speech

POST /v1/audio/speech returns the whole audio file, synchronously.

speech = client.audio.speech.create(
    model="tts-1",
    voice="nova",
    input="Your order is packed and waiting at the pickup point.",
)
speech.write_to_file("out.mp3")

const speech = await client.audio.speech.create({
  model: 'tts-1',
  voice: 'nova',
  input: 'Your order is packed and waiting at the pickup point.',
})
await Bun.write('out.mp3', speech)

curl https://api.mixen.ai/v1/audio/speech \
  -H "Authorization: Bearer $MIXEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "tts-1", "voice": "nova", "input": "Your order is packed and waiting at the pickup point."}' \
  -o out.mp3

Field	Description
`input`	The text, required
`model`	`tts-1` is accepted for compatibility; you can also name a catalog model
`voice`	`alloy`, `echo`, `fable`, `nova`, `shimmer` and others — the full list is the `voices` field in the catalog. Behaviour depends on the model: `gpt-4o-mini-tts` (the default) silently swaps an unknown voice for the default one, `grok-voice-tts-1.0` returns a `400` listing the allowed voices, and `seed-audio-1-0` takes a preset speaker ID (`ru_male_pavel_uranus_bigtts` and more, listed in the `voices` field; without `voice` — the default speaker) ElevenLabs models use premade names (`george`, `sarah`…, 21 voices in the `voices` field)
`response_format`	Audio format
`speed`	Speech rate

Billing depends on the model. Most (gpt-4o-mini-tts, Grok, MiniMax, Voxtral, Qwen, MAI-Voice) charge by the length of the input text, per 1000 characters — the budget is known upfront. Seed Audio and Gemini TTS charge by the duration of the generated audio (seconds for Seed, audio tokens for Gemini): the charge follows the actual output; price units are listed in GET /v1/models: per_1k_chars (characters), per_minute (Seed), per_second (Voxtral/Nemotron), per_hour (MAI), or the input_rub_1m/output_rub_1m pair (Gemini TTS and token-metered transcriptions).

The live TTS line-up is in GET /v1/models, entries with type: tts:

Model	Key	Voices
MiniMax Speech 2.8 HD / Turbo	`minimax/speech-2.8-hd`, `minimax/speech-2.8-turbo`	45 ready-made voices in the `voices` field (`English_Graceful_Lady`…); other MiniMax voice ids work too — model card
Gemini 3.8 Flash TTS / Lite	`google/gemini-3.8-flash-tts`, `google/gemini-3.8-flash-lite-tts`	30 voices in the `voices` field — model card
Gemini 3.1 Flash TTS	`google/gemini-3.1-flash-tts-preview`	Closed roster, in the `voices` field
Grok Voice TTS 1.0	`x-ai/grok-voice-tts-1.0`	Closed roster; an invalid voice returns `400`
Qwen Audio 3.0 TTS Flash / Plus	`qwen/qwen-audio-3.0-tts-flash`, `qwen/qwen-audio-3.0-tts-plus`	Two voices each, in the `voices` field
Voxtral Mini TTS	`mistralai/voxtral-mini-tts-2603`	30 presets in the `voices` field: 4 speakers × emotions (`en_paul_neutral`, `gb_jane_neutral`…); other ids are rejected
Seed Audio 1.0	`bytedance-seed/seed-audio-1-0`	Preset speaker ID in the `voices` field (5 Russian, 12 English); speech and sound effects, up to 120 s
MAI-Voice-2.1 / Flash	`microsoft/mai-voice-2.1`, `microsoft/mai-voice-2.1-flash`	Full voice ID with the model suffix (`en-US-Harper:MAI-Voice-2.1`); 23 languages, ~100 voices
Eleven v4 / v4 Turbo	`elevenlabs/eleven-v4`, `elevenlabs/eleven-v4-turbo`	Premade names (`george`, `sarah`…) — 21 voices in the `voices` field; audio tags like `[whispering]` in the text, up to 10,000 characters
Eleven Multilingual v2	`elevenlabs/eleven-multilingual-v2`	Long-form (voice-overs, audiobooks), 29 languages
Eleven Flash v2.5	`elevenlabs/eleven-flash-v2.5`	Real-time, 32 languages, up to 40,000 characters

The endpoint requires the audio scope — a key without it gets a 403. The opposite direction — speech to text — is covered in the transcription guide.