Black Friday deposit bonus·Top up $50+ and get $10 · Top up $100+ and get $25·Ends November 30Top up →
Free · no sign-up to listen

AI voice library

Hear every text-to-speech voice we host before you write a line of code. Each voice reads the same sentence, so you can compare them side by side, then copy the request that makes it.

$0.036 per minute of audio, billed by the second

12 voices

Sample text: “Hello, and welcome. This is a short sample of my voice, so you can hear how I sound before you pick me for your project.”

Qwen3-TTS

qwen3-tts

Style instructions, voices designed from a description, voice cloning

Languages
10 languages
Price
$0.036 / min

ZONOS2

zonos2

Emotions (happy, sad, angry, surprised), close voice cloning

Languages
40+ languages
Price
$0.036 / min

Pick a voice, then make it yours

These are the preset voices of the two open text-to-speech models we run on our own GPUs, Qwen3-TTS and ZONOS2. Every voice speaks every language its model supports; the language under each name is the one it sounds most natural in.

A preset is only the start. Give a Qwen3-TTS voice style instructions, describe a brand-new voice in words, or clone your own voice from a short recording. ZONOS2 adds emotions. All of it goes through one OpenAI-compatible endpoint with the same API key you use for chat models.

The first request after a quiet spell can take up to two minutes while the model loads; after that, speech comes back faster than real time.

Hear your own script

Open the speech playground, paste your text and pick any voice here. Change the style, the language or the speed, download the audio, or save a clip as a voice of your own.

Frequently asked questions

Is the voice library free?

Yes. Every sample on this page is free to play, with no account. Making your own audio through the API or the playground is billed by the second from a prepaid balance.

How much does text to speech cost?

$0.036 per minute of audio, billed by the second (at least one second per request). There is no subscription and no charge per character.

Can these voices speak other languages?

Yes. Every voice speaks every language its model supports, best in its native one: Qwen3-TTS has 10 languages, ZONOS2 more than 40. Set language in the request, or leave it out and the model detects it.

Can I change how a voice sounds?

On Qwen3-TTS, add instructions such as "slow and calm" or "excited, a little fast" to any voice, or leave the voice out and describe a new one. On ZONOS2, set an emotion: happy, sad, angry or surprised.

Can I clone my own voice?

Yes. Both models clone a voice from 3 to 15 seconds of clear speech. Upload or record it in the playground, or save it through the API, and use it like any voice here. You must have the speaker’s consent.

Does it work with the OpenAI SDK?

Yes. The endpoint is OpenAI’s POST /v1/audio/speech: point the SDK at our base URL with your key. OpenAI voice names such as alloy or nova map to the closest voice here.