# Voices and engines (https://voicelabs.now/docs/voices-and-engines)



## Voice profiles [#voice-profiles]

Your voices live on your account as **voice profiles** — cloned voices and presets. Each profile carries a name, a language, the TTS engine it uses, and usage counts. When you generate speech (in the studio, through the MCP server, or through the API), you pick the profile to speak in.

## TTS engines [#tts-engines]

Seven text-to-speech engines, including **Qwen3-TTS**, **Chatterbox**, and **Kokoro**, so you can pick the model that fits your quality, language, and speed needs on every generation. All inference runs on VoiceLabs' GPU servers — your browser never needs a GPU.

<Callout title="Engines change over time">
  The lineup above is what runs today. VoiceLabs posts real, dated release notes on the [changelog](https://app.voicelabs.now/changelog) whenever an engine is added, replaced or retired.
</Callout>

## Voice cloning [#voice-cloning]

Instant voice cloning from a short reference clip, in seconds. Two rules apply from day one:

* You may only clone a voice from a sample you are authorized to use.
* The [Responsible Use policy](https://app.voicelabs.now/responsible-use) applies to every generated voice.

## AI-generated audio disclosure [#ai-generated-audio-disclosure]

Audio created with VoiceLabs is synthetic. Where the law or a platform requires it, disclose it as AI-generated. VoiceLabs marks exported audio machine-readably toward EU AI Act Article 50 (in force August 2, 2026).