Can ChatGPT transcribe audio and video?

Audio, yes, on paid plans. Video, no. You can upload an audio file to a chat or record a meeting with Record mode on a Mac. You get plain text back: no speaker labels you can rely on, no timestamps, no subtitle file. The free microphone button is dictation, not transcription.

By VexaScribe Editorial · Published August 15, 2026 · Rewritten November 17, 2026 · OpenAI facts checked November 17, 2026

Short answer

  • Free: dictation and voice chat, but no file transcription. You can paste a transcript you made elsewhere and ask ChatGPT to clean it up or summarise it.
  • Paid plans: upload MP3, M4A, WAV and other audio files, up to 512 MB.
  • Mac desktop app on Plus and above: Record mode, up to 4 hours per session.
  • Video: not supported. Extract the audio first.

The ways ChatGPT handles audio

MethodPlansWhat it doesCatch
Dictation (microphone button)AllTurns your speech into the message you're typingCan't open a file
Voice conversationAll, with daily limits on FreeSpoken back-and-forth with ChatGPTNot a transcript of a recording
Upload an audio filePaid plansTranscribes or summarises the file in the chatNo timestamps or subtitle files
Record modePlus, Pro, Business, Enterprise, EduRecords a meeting live; gives a transcript and notesmacOS desktop app only; can't open existing files
OpenAI APIDeveloper accountTranscribes files from your own code25 MB per request; you write the code

The mix-up people hit most: the microphone in the message box is dictation. It listens to you, live, and types your words into the chat. There's no way to point it at a recording.

Transcribing an audio file

  1. Open a chat on a paid plan and attach the file with the paperclip.
  2. Ask for what you want: "Transcribe this word for word" or "Summarise this meeting and list the action items".
  3. Read the result against the audio, especially names, numbers and anything you'll quote.

OpenAI accepts WAV, MP3, M4A, FLAC, OGG, AAC and audio-only MP4 or WebM, up to 512 MB. It doesn't promise a complete word-for-word transcript of a long recording, so check that nothing is missing from the middle or end.

Record mode

Record captures a meeting, lecture or voice note live and turns it into a transcript plus a notes canvas you can edit. It's for new recordings: it can't open a file you already have.

  • Plans: Plus, Pro, Business, Enterprise and Edu; off by default on Enterprise and Edu until an admin enables it
  • Where: the ChatGPT desktop app for macOS. Not on Windows, the web or phones.
  • Length: up to 4 hours per session, after which it stops and saves the notes
  • Audio: deleted after it's transcribed; the transcript stays in your chat history
Can't find Record on your Mac? Users report it missing from some newer versions of the desktop app. Check OpenAI's Record help page for the current requirements.

Video files

ChatGPT's help page is explicit: MP4 and WebM files identified as video aren't supported for audio. It can't play a YouTube link either. Extract the audio, then upload that:

FFmpeg: keep the audio track from a video
ffmpeg -i video.mp4 -vn -c:a copy audio.m4a

Or skip the step: dedicated transcription tools, ours included, take the video file as it is.

What ChatGPT doesn't give you

  • Timestamps. You get prose without times, so you can't jump to a moment or make subtitles.
  • Subtitle files. No SRT or VTT. If you ask, it will write times that weren't measured from the audio.
  • Reliable speaker labels. Asking ChatGPT to split a transcript by speaker afterwards is a guess based on the words, not the voices.
  • A saved, searchable library. Each transcript lives in a chat.

Where ChatGPT shines is what comes after: summarising, translating, pulling out action items or rewriting a transcript, all in the same chat.

Prompts that work and fail

Work well, once you have a transcript

  • "Summarise this in five bullet points, then list decisions and action items with owners."
  • "Remove filler words and false starts, but don't change the meaning or reorder anything."
  • "Pull out every question the audience asked, with the answer given."
  • "Translate this into Spanish, keeping names and numbers exactly as they are."

Fail, and what to do instead

  • "Transcribe this YouTube video" with a link: ChatGPT can't play it. Download the audio or use a tool that imports the video's captions.
  • "Fix the timing in this SRT": it can't hear the audio, so new times are invented. Use a subtitle editor such as our SRT editor.
  • "Put back the ums and ahs": words that weren't transcribed can't be restored from the text. Transcribe again in a verbatim style.

The OpenAI API

For developers, OpenAI's recommended model for recorded files is gpt-transcribe, at $0.0045 a minute on November 17, 2026 (see OpenAI pricing). Files are limited to 25 MB per request, so long recordings need splitting.

One file, from the command line
curl https://api.openai.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -F model="gpt-transcribe" \
  -F file="@meeting.mp3"
Older models are going away. whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize were deprecated on August 26, 2026 and are removed on February 26, 2027. OpenAI's guide still documents word timestamps and subtitle output only for whisper-1, and speaker labels only for the diarize model, so check the replacements before you migrate.

Need speakers, timestamps or subtitles?

Upload audio or video here, video included. You get speakers, timestamps and an SRT download for the first 5 minutes, free with no account.

Drop your audio file here

or click to browse

Audio: MP3 · WAV · M4A · AAC · OGG · OPUS · WMA · FLAC · AMR · AIFF

Video: MP4 · MOV · MPEG · WMV · AVI · MKV · WEBM · FLV · 3GP · M4V

Up to 200 MB · first 5 minutes free, no account

Sign up free for 30 minutes and files up to 5 GB, no card.

Sources

OpenAI changes plans and limits often; checked November 17, 2026. VexaScribe is independent and not affiliated with OpenAI. See our editorial standards.

Frequently Asked Questions

Can ChatGPT transcribe audio for free?

No. The free plan has dictation and voice chat, which work on your live voice, but uploading an audio file needs a paid plan. On Free you can paste a transcript made elsewhere and ask ChatGPT to clean it up or summarise it.

How big an audio file can I upload to ChatGPT?

Up to 512 MB, in formats including MP3, M4A, WAV, FLAC, OGG, AAC and audio-only MP4 or WebM, on paid plans. Long recordings may not come back complete, so check the end of the transcript.

Can ChatGPT transcribe a video?

No. OpenAI says MP4 and WebM files identified as video aren't supported for audio, and ChatGPT can't play a YouTube link. Extract the audio first, for example with FFmpeg, or use a transcription tool that accepts video.

Does ChatGPT give timestamps or SRT subtitles?

No. Uploads and Record mode return text without timestamps. If you ask for an SRT, ChatGPT writes times that weren't measured from the audio. Use a transcription tool that exports SRT, then use ChatGPT on the text.

Where is ChatGPT Record mode available?

In the ChatGPT desktop app for macOS, on Plus, Pro, Business, Enterprise and Edu. Sessions last up to 4 hours. It isn't available on Windows, the web or phones.

Also available in: Deutsch