ChatGPT adds audio file uploads for transcription, summaries and questions

OpenAI added support for uploading audio files to ChatGPT on October 6, 2026. Subscribers can attach an existing recording, request a transcript or summary, and ask follow-up questions about what was said. The update is useful for lectures, interviews, meetings and other recorded conversations.

From recorded audio to searchable notes

Users can attach a supported audio file and ask ChatGPT to transcribe it, identify key topics, organize notes or prepare an action-item recap. The feature analyzes recordings that users already have; it should not be confused with a new built-in feature that automatically records a 90-minute meeting.

OpenAI lists supported formats including MP3, WAV, M4A, AAC, FLAC and OGG. Audio uploads are limited to 512 MB per file. Long recordings may be divided into smaller sections when Data Analysis is available, but processing is best-effort and especially long files may time out.

U.S. availability and pricing

Audio uploads are available through paid ChatGPT subscriptions and supported workspaces, including eligible enterprise deployments. They are not currently included in the Free plan. OpenAI has not announced a separate per-audio-file price for this ChatGPT feature.

Availability may differ by region, account plan, selected model, workspace configuration and app version. U.S. users who do not see the option should check their account and client availability rather than assume that every paid workspace has it enabled by default.

Accuracy and privacy considerations

Transcripts can contain mistakes, and speaker identification may be unreliable. Background noise, overlapping speech, technical vocabulary and language differences can affect results. For legal, academic or business use, important quotations and conclusions should be checked against the original recording.

Users should also confirm that they are permitted to record or upload the conversation and avoid sharing sensitive audio without appropriate authorization.

KX3 perspective: capture quality still matters

Audio analysis has become easier to access, but the quality of the result still depends heavily on the source recording. Microphone placement, background noise and reliable long-duration capture remain crucial. A practical workflow can separate recording from downstream AI transcription, allowing users to choose the most suitable tool for each stage.

For organizations with strict data-location requirements, locally processed transcription remains an alternative worth evaluating alongside cloud uploads.

Announcement and sources

Announcement: October 6, 2026. Retrospective archive article prepared October 9, 2026.

Primary sources: OpenAI ChatGPT release notes; OpenAI audio-upload documentation.

Illustrative photo: Jonathan Velasquez / Unsplash (not an official ChatGPT screenshot).

日本語版を読む