OpenAI makes prompt-to-voice generally available in the API: describe a voice in text to create a custom GPT-Live voice for free, plus nine new native-speaker voices in five languages

Why it matters

【Key points】

OpenAI's developer changelog (Oct 16) says prompt-to-voice for GPT-Live is now generally available to all API customers. Developers describe the voice they want in text and create a saved custom voice through the API or the voice Playground, then reuse it in GPT-Live sessions. Creating the voice is free; speech is billed as normal GPT-Live session usage. @OpenAIDevs promoted it on Oct 16 at 17:08 UTC.

OpenAI also added nine GPT-Live preset voices recorded with native-speaking actors: Nuri, Juni and Flitz (German), Brise and Sillage (French), Nira (Hindi), Harema and Shida (Japanese) and Noeul (Korean).

Documented limits: prompt-generated voices work only with GPT-Live, not the Text to Speech, Realtime or Chat Completions APIs. Voices cloned from a recording still need a consent recording plus a reference sample. The usage rules ban impersonation and children's voices and require telling listeners the speech is AI-generated.

【Verifiable】

Changelog: https://developers.openai.com/changelog

Custom voices guide: https://developers.openai.com/api/docs/guides/custom-voices

@OpenAIDevs post: https://x.com/OpenAIDevs/status/2111142258867335430

【Our read】

1. This is the API-side follow-up to GPT-Live-1.1 reaching ChatGPT (e/893). Developers can now design a branded voice from a description instead of picking a preset or recording a speaker, which makes voice agents cheaper to build.

2. Voices made without a recording carry less impersonation risk than cloning, but the real test is how OpenAI enforces its no-impersonation and disclosure rules.

3. The nine new voices cover German, French, Hindi, Japanese and Korean, a sign OpenAI is working on voice quality outside English.