OpenAI has upgraded ChatGPT's dictation feature into a real-time transcription tool: words now appear on screen as you speak, rather than only after the recording has been processed. The company says transcription is faster and more reliable, and it has removed the previous 10-minute cap on dictation recordings entirely.
The old workflow made long-form voice input awkward. A user dictating a lengthy prompt or a set of notes saw nothing until processing completed, and anything past ten minutes simply hit a hard limit. With Live Dictation, the transcript streams into view while you talk, which makes it possible to catch a misheard word immediately, rephrase mid-sentence, or deliver long brainstorms and multi-step instructions in one continuous take.
The feature is separate from ChatGPT's conversational Voice mode, which handles spoken back-and-forth dialogue. Live Dictation is built for getting text into the prompt box: the final transcript lands there for editing before sending, and users can toggle the live transcript view on or off by tapping the audio waveform while speaking. OpenAI confirmed the changes publicly; coverage by 9to5Mac and The Verge notes OpenAI's Atty Eleti confirmed the removal of the time limit and the reliability improvements.
The rollout is global and broad: all ChatGPT plans, including the free tier, across iOS, Android, the web and desktop apps, reaching accounts progressively. OpenAI did not publish latency or accuracy benchmarks and has not stated a new maximum session length — only that the feature is faster and more dependable than before. Users who prefer not to share audio can manage dictation-related sharing options, including the option that covers recordings used for model improvement, in ChatGPT's data controls.
The upgrade slots into a voice push that has been building all year. Earlier in 2026 OpenAI overhauled ChatGPT's two-way voice conversation mode, and its beta record mode for transcribing meetings and voice notes — currently limited to Plus, Enterprise, Edu, Business and Pro workspaces — continues to run alongside the new dictation path. Third-party coverage of the rollout has cited weekly users of ChatGPT voice and dictation features reaching roughly 250 million by early November, up from about 150 million in July, though OpenAI has not included that figure in its own announcement.
Why it matters: dictation is quietly becoming the default interface for a growing class of users — writers thinking out loud, developers issuing long instructions, accessibility users — and live streaming transcription matches what dedicated tools like Wispr Flow already do. Bundling it into ChatGPT on every plan, free included, turns voice input from a niche power feature into table stakes, and puts pressure on standalone dictation apps whose main selling point was exactly this experience.
Comments (0)
Log in to join the discussion
Log InNo comments yet