OpenAI to Retire ChatGPT's Legacy Voice Modes on Dec 11, Leaving GPT-Live Only

What happened

ChatGPT's voice feature has offered three modes since Advanced Voice launched in 2024: Standard Voice, which turns speech into text before answering; Advanced Voice, which processes audio directly; and GPT-Live, a full-duplex model introduced in July 2026. The two older modes take turns speaking, while GPT-Live listens and speaks at the same time and can hand complex queries to frontier text models without interrupting the conversation.

On Nov 10, OpenAI began notifying users that Advanced and Standard Voice will disappear from Settings — Voice on Dec 11, and anyone still on the older modes will be switched to GPT-Live automatically; past voice conversations will remain in history and can continue by text or with GPT-Live. OpenAI said voice usage in ChatGPT has nearly doubled since GPT-Live launched in July, and most voice users have already moved to it.

After the announcement, some users on X and Reddit complained that GPT-Live interrupts often, gives shorter answers and holds continuity worse than the older modes, and some paying subscribers said they might cancel. OpenAI acknowledged that some people prefer the older modes' response style and said it will use what it learned from them to improve GPT-Live's response depth, conversational continuity and personalization. In September 2025, OpenAI said Standard Voice would be retired within 30 days, then kept it after users pushed back and ChatGPT head Nick Turley said so on X; this time the company put a fixed cutoff in its release notes.

Key facts

1. Scope: Advanced and Standard Voice retire on Dec 11; GPT-Live becomes ChatGPT's only voice mode
2. Timeline: Notifications began Nov 10; the older modes remain usable until Dec 11, when they disappear from settings and their users switch to GPT-Live automatically
3. Exceptions: Live video and screen sharing keep using Advanced Voice, as GPT-Live does not support them yet; voice chats with custom GPTs keep using Advanced Voice until Dec 11
4. GPT-Live: Launched in July 2026 with a full-duplex architecture, it listens and speaks at the same time and can hand complex queries to frontier text models without interrupting the conversation
5. Scale of the migration: OpenAI said voice usage nearly doubled since GPT-Live launched in July, and most voice users have already moved to it
6. Precedent: In September 2025, OpenAI said Standard Voice would go within 30 days, then kept it after users pushed back; this time it set a fixed Dec 11 cutoff