Can ChatGPT Transcribe Audio? Yes, Paid Plans, 6 Min Cap

In brief

Yes, on a paid plan. ChatGPT has transcribed uploaded audio since 6 October 2026, but only recordings up to 6 minutes long in Chat, and OpenAI says audio uploads are not available on the Free plan. Free users can still dictate by microphone. Record mode on the Mac app runs up to 240 minutes.

Can ChatGPT transcribe audio, and on which plan?

ChatGPT transcribes audio three ways, and each has its own plan and limit. OpenAI's uploading files and audio article says it flatly: "Audio uploads are not available on the Free plan at this time." Paid plans get uploads. Every plan gets dictation. The table sorts the routes as of November 2026, from OpenAI's help center and API docs.

A brushed silver OpenAI knot on a black plinth beside a small studio microphone and a short curl of blank paper tape, and a rainbow glass Gemini star on a low plinth beside a long coil of paper tape and a pair of headphones
RouteWho gets itLimitWhat you get
Upload an audio file in ChatPaid plans and workspaces, not Free6 minutes per recording on GPT-6 at Instant, up to 512 MBTranscript, summary or answers about the recording
Record modePlus, Enterprise, Edu, Business and Pro workspaces, macOS desktop app only240 minutes per sessionLive transcript and a summary saved as a canvas
Dictation (microphone icon)All plansRecords a spoken messageText you can edit before you send
Speech to text APIDevelopers25 MB per filegpt-transcribe, built for recorded speech
ChatGPT uploads stop at 6 minutes, Record mode runs 40 times longer
5 min20 min1 hour4 hoursChatGPT audio uploadGemini, no planGemini on AI ProChatGPT Record mode

The chart puts the longest recording each route takes on one line. ChatGPT's upload cap is the shortest of the four. Gemini's file upload page says "Total audio length can be up to 10 minutes," and a Google AI Pro or Ultra plan raises that to 3 hours. Google's separate Gemini Notebook also transcribes a local audio file on import, but its help page states no length cap. Claude's upload article lists documents and four image formats and no audio, so there is nothing to compare there.

How do you transcribe an audio file in ChatGPT?

Attach the file to a chat and tell ChatGPT what output you want. OpenAI's help page gives exactly that recipe: attach a supported file, describe the result, then review and ask follow-ups. It accepts WAV, MP3, OGG, audio only WebM, PCM, FLAC, AAC, M4A and audio only MP4, with the file up to 512 MB and a valid, decodable audio stream.

Copy this prompt and change the brackets:

Transcription prompt: "Transcribe the attached recording word for word. Label each speaker as Speaker 1, Speaker 2 and so on. Mark any passage you could not hear clearly as [unclear] with the time. After the transcript, list the decisions made and every action item with the person who owns it. The language is [English]."

Two things help. Ask for [unclear] markers, because a transcript that guesses is worse than one that admits a gap. And name the language, since OpenAI says quality may vary across languages.

ChatGPT's Work and Codex modes do not take uploaded audio for transcription, per the same page. Open a normal Chat. Availability also depends on workspace settings, region, client version and the model you picked.

Why won't ChatGPT transcribe my audio file?

Most failures trace to five causes, in this order of likelihood. OpenAI notes that a file can attach or preview fine even when transcription is not available, so a successful upload proves nothing.

SymptomLikely causeFix
Upload is refused or ignoredYou are on the Free planUpgrade, or dictate by microphone
Transcript stops early or failsRecording is over 6 minutesSplit it into parts under 6 minutes
File is rejected as videoA WebM or MP4 file is marked as videoExport audio only, or convert to WAV
File attaches, no transcriptNot a decodable audio streamConvert to WAV and upload again
No transcription option at allYou are in Work or CodexStart a plain Chat

OpenAI's own troubleshooting tip is the WAV conversion: if a recording uploads but won't transcribe, convert it to WAV and try again.

What can you use when the recording is longer than 6 minutes?

Pick by length. For a meeting or lecture on a Mac, use ChatGPT's Record mode. OpenAI's record article says ChatGPT "can transcribe and summarize audio recordings like meetings, brainstorms, or voice notes," labels speakers as Speaker 1 and so on, and caps a session at 4 hours. Audio files are deleted after transcription. It is Mac only. OpenAI's page lists Plus, Enterprise, Edu, Business and Pro workspaces in one note and calls it a beta for Pro and Business in another, so check your own app.

For a file you already have, Gemini takes 10 minutes without a plan and 3 hours on Google AI Pro or Ultra, per Google. For a developer, OpenAI's speech to text guide recommends gpt-transcribe for recorded speech and allows files up to 25 MB, in flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav and webm. The older whisper-1, gpt-4o-transcribe and gpt-4o-mini-transcribe shut down on 26 February 2027. The API takes a file, not a link, per OpenAI's Audio API FAQ.

The transcript is the start, not the end. Most people then need minutes, a summary or a clean rewrite, and the best model for that varies by job. Whizi Pro carries Claude, GPT and Gemini in one app, with a first week at $1.99, so you can run one transcript through each and keep the write up you like best. The free plan gives 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5, no card. Whizi's voice mode is a live spoken conversation, not a file transcriber, so make the transcript first.

How accurate is ChatGPT's transcription?

OpenAI gives no accuracy figure for uploads and says to check the output. Its help page warns that "Transcripts may contain errors" and that speaker identification may be unreliable. Treat names, numbers and quotes as unverified until you replay them.

Dictation has one published number. OpenAI's release notes say a new speech to text model rolled out to all plans on 26 June 2026, with word error rate at least 10% lower for the top languages tested than the previous production model. That figure compares OpenAI's two models, not ChatGPT against a human typist.

Privacy differs by route. For dictation, OpenAI's voice dictation FAQ says audio "will be retained for as long as the chat is part of your chat history," and consumers who opt in may have audio used for training. Record mode deletes audio files after transcription. For a confidential interview, read both pages before you upload.

Workflow checklist
  • Confirm you are on a paid plan before you upload
  • Cut each recording to 6 minutes or less
  • Use MP3, WAV or M4A, not a video file
  • Ask for [unclear] markers and name the language
  • Replay every name, number and quote before you use it
Common questions

Frequently asked questions

Can ChatGPT transcribe audio for free?

Only by dictation. OpenAI says audio uploads are not available on the Free plan, so a free account can't upload a file for a transcript. The microphone icon turns one spoken prompt into text on every plan, and a new dictation model reached all plans on 26 June 2026. ChatGPT Free vs paid lists what else the free plan lacks.

Can ChatGPT transcribe audio from a video or a YouTube link?

Not from a link. ChatGPT takes an attached file, and uploads that OpenAI identifies as video, WebM or MP4, are not supported by audio uploads. Export the soundtrack as MP3 or WAV first. OpenAI's Audio API FAQ adds that its API needs a file in a supported format, not a link.

Can ChatGPT transcribe audio in different languages?

Yes, with uneven quality. OpenAI says transcription quality may vary across languages, and its older Whisper model supports 98 languages with accuracy that varies by language. Name the language in your prompt and check the output in any language you can read.

Can ChatGPT transcribe a live meeting?

Yes, with Record mode on the Mac app. It live transcribes while a timer runs, labels speakers as Speaker 1 and so on, and saves a summary as a canvas. Sessions stop at 240 minutes. It isn't on the Free plan. ChatGPT voice mode covers the separate spoken conversation feature.

Can ChatGPT transcribe a voice memo or an MP3?

Yes, on a paid plan, if the recording is 6 minutes or shorter. OpenAI lists MP3 and M4A among the formats audio uploads accept, with files up to 512 MB. On the Free plan the upload is refused, so dictate into the microphone icon instead.