OpenAI Kills Standard and Advanced Voice Modes by December 11

On December 11, 2026, OpenAI will remove both the Standard Voice Mode and the Advanced Voice Mode from ChatGPT's speech settings, migrating every user automatically to its newer real-time system, GPT-Live. The company began notifying users about the shutdown on November 10, 2026 — a month's notice for a change that retires two generations of speech architecture at once.

Existing voice conversations won't vanish. OpenAI says past voice chats stay stored in the chat history, and users can resume those archived threads either by typing text or by continuing them inside GPT-Live.

What changes technically?

The three systems differ in how they handle speech, and the differences explain why OpenAI wants to consolidate. Standard Voice worked in three discrete steps: it transcribed the spoken word to text, generated a written reply from that transcript, then read the reply aloud. The Advanced Voice Mode processed audio signals directly, but users still noticed noticeable pauses between turns.

GPT-Live, which launched on July 8, behaves differently again. It listens and speaks simultaneously, tolerates interruptions when the user asks a mid-sentence question, and hands complex tasks to a text model running in the background. In conversational terms, it behaves less like a walkie-talkie exchange and more like a phone call.

Where does GPT-Live fall short?

The new system doesn't cover everything yet. Live video and screen sharing remain unsupported in GPT-Live, and OpenAI has named no date for integrating them. Until December 11, those features — and voice conversations with customized variants, the so-called Custom GPTs — still rely on the Advanced Voice Mode.

The reaction from parts of the user base has been prickly. On Reddit and X, some users complained about the retirement of the two older modes. Free-plan users face an additional constraint: GPT-Live is available to them only in a limited form.

OpenAI justifies the shift by pointing to growth and engagement. According to the company, use of the voice functions has nearly doubled since GPT-Live launched on July 8, and nearly all active voice users have already switched to the newer system. The company says user feedback will feed continuously into the live features, with the stated goals of deeper answers, better dialogue continuity, and stronger personalization.

How did the voice lineup grow this autumn?

The migration caps weeks of steady expansion. OpenAI first added two Hindi voices, named Leher and Neer. On October 15, 2026, it shipped updates to GPT-Live-1.1 and GPT-Live-1.1 Quip, which added voices in Korean among other changes. Later in October 2026, two Portuguese voices joined the lineup, under the names Viola and Rio.

What happens to stored audio?

The data retention rules differ by mode, and the older settings don't all share one fate:

• Audio recordings from Live and Advanced sessions, plus video content from the Advanced mode, are kept for 30 days.
• Audio from the Standard mode is deleted immediately after transcription.
• Data export requests take up to seven days to process, and each download link stays valid for 24 hours.

Why the timing matters

The consolidation lands against a backdrop of market dominance. As of September, ChatGPT accounted for 81.2 percent of global AI chatbot sessions, across 61.35 billion monthly sessions and roughly 1.2 billion weekly active users. When a platform of that scale flips a default, the interface habits of a billion people change overnight — including free-tier users who will meet the new system with restricted access and no legacy option to fall back on after December 11.