On October 15, OpenAI added native Decisions integration to GPT-Live. An application can act while someone speaks or respond to what Live says. The announcement describes two paths: let Live delegate directly to Decisions, or trigger decisions as incoming or outgoing transcript deltas arrive.
On October 16, prompt-to-voice became available to all API customers: a text description creates a voice through the API or Playground for reuse in GPT-Live sessions. Voice creation is free, but speech generation uses normal session billing. Nine natural voices were also added: Nuri, Juni and Flitz in German; Brise and Sillage in French; Nira in Hindi; Harema and Shida in Japanese; and Noeul in Korean.
These announcements concern API applications, not the GPT-Live-1.1 rollout in ChatGPT Voice announced on October 15. The distinction matters for both access and billing: a voice that is free to create does not make voice conversations free.
Prompt-to-voice: all API customers according to OpenAI. No specific Belgian exclusion is stated; API eligibility and supported-country rules still apply.
ChatGPT availability and voice options do not establish API access. Accent quality and action safety have not been independently measured.