9 voices mentioned in official openai API reference gets rejected by the API

Environment

  • Surface: API
  • Endpoint or feature: /v1/live/sessions
  • Model: gpt-live-1
  • SDK and version, if applicable: OpenAI’s latest Node SDK (7.27.0)

Bug:

brise, flitz, harema, juni, nira, noeul, nuri, shitan, sillage that are mentioned here, gets rejected with following error, while other 22 names are accepted: (API reference for above endpoint)

https://developers.openai.com/api/reference/resources/live/methods/create?site_locale=en#(resource)%20live%20%3E%20(method)%20create%20%3E%20(params)%200%20%3E%20(param)%20session%20%3E%20(schema)%20%2B%20(resource)%20live%20%3E%20(model)%20media_session_config%20%3E%20(schema)%20%3E%20(property)%20audio%20%3E%20(property)%20output%20%3E%20(property)%20voice%20%3E%20(variant)%201

403 forbidden “Voice session access denied.”

Thanks for reporting this, @dhyey.searce.

I’ve been able to reproduce the issue and have flagged it with the team for investigation. We’ll share updates in this thread as they become available.

Hi @dhyey.searce, sorry for the wait. We reproduced the same 403s with direct API requests too, so this isn’t specific to the Node SDK. The team is tracking it. For now, you can use one of the voices that worked in your tests.

@OpenAI_Support sure, it’s not node SDK specific but please update when those voices can be supported, I’d love to try them out.