Can ChatGPT Transcribe Audio or Video? What Works in 2026

Partly. According to OpenAI's help center (checked October 2026), ChatGPT accepts audio uploads such as MP3, M4A and WAV on paid plans, not on the Free plan, and files detected as video aren't supported for audio transcription. For a complete word-for-word transcript of a long recording or a video, a dedicated transcription tool is more reliable.

Short answer: on a paid ChatGPT plan you can upload an audio file and ask for a transcript. On the Free plan you can't, and an MP4 or WebM that ChatGPT identifies as video isn't accepted for audio transcription. If you need every word of a long meeting, lecture or video, transcribe the file first, then paste the text into ChatGPT to summarize it.

What it can and can't transcribe

OpenAI's help center describes audio uploads as a feature of paid ChatGPT subscriptions and workspaces. The formats it lists include WAV, MP3, M4A, FLAC, AAC and OGG, plus MP4 or WebM files that contain only audio. It also warns that transcripts may contain errors, that quality varies between languages, and that availability can depend on workspace settings, region, app version and the model you pick.

In practice, that means:

• It can transcribe a voice memo, a podcast clip or an interview saved as MP3, M4A or WAV, on a paid plan.
• It can't take any audio file on the Free plan.
• It can't transcribe a video file through audio uploads (more on that below).
• Don't expect subtitles. The transcript comes back as chat text. If you ask for SRT, check every timing against the audio, because subtitles only work when the times match the speech.

OpenAI's help center also describes record mode in the ChatGPT desktop app for macOS, on some paid plans. It records a meeting while it happens and turns it into notes. That helps with the next call, not with a file you already have.

Here is how ChatGPT compares with a tool built only for transcription:

TranscribTxt is a transcription tool only: it doesn't label speakers, doesn't translate, and doesn't summarize. That's what the second step in ChatGPT is for.

Why long files come back cut or summarized

ChatGPT is a chat assistant that can listen, not a transcription program. Three things get in the way of a long, exact transcript:

1. A reply has a limited length. People speak at roughly 130 to 150 words a minute, so one hour of a meeting is about 8,000 to 9,000 words. A reply that long may stop early or come in parts, and it's easy to miss that the last part never arrived.
2. The assistant is tuned to be helpful, not literal. Ask it to "transcribe this" and you may get cleaned-up sentences, a shortened version, or a summary with headings. That's fine for notes and risky for quotes.
3. Errors are possible, and OpenAI says so. Like any speech AI, it can mishear a name or fill a noisy gap with words nobody said. Our guide on why AI transcription makes up words explains when that happens.

If you do use ChatGPT, ask for it plainly: "Transcribe this audio verbatim, word for word. Don't summarize, don't fix grammar, don't skip anything. If you run out of space, stop and tell me the last sentence you transcribed." Then compare that sentence with the end of the recording.

A video ChatGPT won't take

Many recordings people want transcribed are videos: a Zoom call, a lecture, a screen recording, a phone clip. According to OpenAI's help center, MP4 and WebM files identified as video aren't supported by audio uploads, even though the same formats work when they hold only audio.

You have two ways around it:

• Extract the audio first. Some apps already save an audio-only copy (Zoom writes an M4A next to its MP4). With the free ffmpeg tool, ffmpeg -i meeting.mp4 -vn -c:a copy meeting.m4a copies the sound track out of most MP4 files without re-encoding it. Then upload the M4A to ChatGPT.
• Skip the conversion. Upload the video itself on our video to text page. MP4, MOV, MKV, WebM and AVI are accepted up to 2 GB, and you download the transcript as TXT or SRT.

Full transcript first, then summarize in ChatGPT

This way you get both: a complete transcript you can check, and a ChatGPT summary based on it.

1. Upload the recording in the box above, or on the audio to text page. Audio and video both work, and the language is detected automatically from 99 supported languages.
2. Wait for the transcript. In October 2026, a 145-minute recording took 2 min 45 s once uploaded.
3. Download the TXT, and the SRT too if you need captions or exact times for quotes.
4. Paste the text into ChatGPT and ask, for example: "Summarize this meeting transcript: decisions, action items with owners, open questions." Pasting text works on any ChatGPT plan, including Free. Paste a very long transcript in parts.
5. Check quotes and numbers against the TXT, not against the summary.

Free covers files up to 30 minutes: the first file needs no account, and a free account gets 3 files a day. A longer recording gets its first 30 minutes free. The whole file is $5 once with a 7-day Week Pass (up to 25 hours of audio, no renewal), or Pro at $10 a month. The uploaded file is deleted after processing (a free-preview upload is kept up to 24 hours so you can finish it), and the transcript isn't stored on our servers, so download it before you leave.

ChatGPT, Gemini, Claude or a tool?

All three assistants are good at working with a transcript. They differ in whether they'll take the recording itself.

• ChatGPT: audio files on paid plans, as above. Video files aren't accepted for audio transcription.
• Gemini: according to Google's Gemini Apps Help (checked October 2026), uploaded audio can total up to 10 minutes without an upgrade and up to 3 hours with Google AI Pro or Ultra, with audio files up to 100 MB. Videos can be up to 2 GB, but their total length is capped at 5 minutes, or 1 hour on those paid plans.
• Claude: according to Anthropic's Claude help center (checked October 2026), uploads cover documents such as PDF, DOCX and TXT, plus images. Audio and video formats aren't listed, so you transcribe first and paste the text.
• A transcription tool: built for the whole file, including long recordings, and it gives you a text file you can keep. Use it when the exact words matter: interviews, lectures, meetings, subtitles.

A simple rule: if you need every word, start with a transcription tool. If you only need the main points of a short clip and already pay for ChatGPT or Gemini, the assistant may be enough. For more options, see our guides to free transcription software, Whisper alternatives for running a model on your own computer, and how to transcribe audio to text.

How we checked (October 2026): we checked OpenAI's, Google's and Anthropic's help pages for file and audio uploads on 26 October 2026 and quote only what they say, with links. TranscribTxt's limits and prices on this page come from our own plan settings.