Can ChatGPT Transcribe Audio?
By Eman Khoubian, Founder & CEO · Updated
Yes, if you pay for ChatGPT and the recording is short. With GPT-6 Instant, chat transcribes uploaded recordings up to 6 minutes long, and anything longer is best-effort and may time out. For an interview or a meeting recording, use WhisperAI's app inside ChatGPT: it takes files as big as 5 GB and sends back a transcript with speaker labels.
How many minutes of audio can ChatGPT transcribe?
Six minutes per uploaded recording, in chat. That's a voice memo, not a meeting, and it's OpenAI's own limit for GPT-6 Instant. The file itself can be up to 512 MB, so a long recording uploads without complaint. The transcript is where it breaks down: past 6 minutes, ChatGPT may work through the audio in smaller sections if its Data Analysis tool is available, and very long recordings may time out.
Here's what each route does with the same file: a 2-hour, 300 MB interview with two people talking.
| Route | What happens to the 2-hour file | Speaker labels | Who gets it |
|---|---|---|---|
| WhisperAI's app in ChatGPT | Uploads the whole file and transcribes all 2 hours. | Yes, Auto or 2 to 10 speakers, and you can name them | Every WhisperAI plan, within its minutes |
| Upload in ChatGPT chat | Uploads fine (the cap is 512 MB), but with GPT-6 Instant, chat transcribes recordings up to 6 minutes long. The rest is best-effort and may time out. | OpenAI says speaker identification may be unreliable | Paid ChatGPT plans, not Free, Work or Codex |
| ChatGPT Record | Can't take it. Record captures a live session on a Mac, up to 4 hours. | Generic labels such as Speaker 1, which you can rename | Plus, Pro, Business, Enterprise and Edu, in the macOS desktop app |
| OpenAI's API | Takes files up to 25 MB, so the 300 MB file goes in as at least 12 pieces, and you write the code. | Only in an older model OpenAI has deprecated | Developers with an API key, paid per minute |
How much of a 2-hour, 300 MB interview each route transcribes in one go
WhisperAI's app in ChatGPT: all 120 minutes as one file. ChatGPT chat upload: 6 minutes promised, the other 114 best-effort. OpenAI's API: all 120 minutes, but as at least 12 separate files of 25 MB. ChatGPT Record: none, because it records live and doesn't take a file.
- WhisperAI app in ChatGPTAll 120 min, one file
- ChatGPT chat upload6 min, rest best-effort
- OpenAI's API12+ pieces of 25 MB
- ChatGPT RecordLive only, not a file
Record is for a meeting that hasn't happened yet, in the Mac desktop app. It runs up to 4 hours a session, works best in English, and deletes the audio once the transcript is made.
The API is for developers. At 25 MB per file, a long recording gets compressed or cut up first and stitched back together after.
How to transcribe audio in ChatGPT
For a voice memo or a short clip, ChatGPT on its own is enough.
- Check the format. Chat takes MP3, M4A, WAV, AAC, FLAC, OGG and PCM, plus MP4 and WebM files that hold audio only.
- Attach the file to a message.
- Ask for the full text before anything else, for example: "Transcribe this word for word and start a new line each time the speaker changes." Ask for the summary in a second message, so the summary doesn't replace the transcript.
- Check names, numbers and technical terms against the audio. OpenAI says its transcripts may contain errors.
The microphone in the message box is dictation, not transcription. It turns what you say into text you can edit before you send it, which is handy for your own words and no use for a recording someone else made.
How to transcribe a long recording and use it in ChatGPT
For an interview or a meeting, let WhisperAI make the transcript and ChatGPT do the thinking. WhisperAI is an independent transcription service, not made by OpenAI, and its app is approved in ChatGPT's app directory, so you never leave the chat.
A long recording, from file to answer, without leaving ChatGPT
Four steps: connect WhisperAI and sign in, open the uploader and set the language and speakers, choose the recording, then ask your questions about the transcript.
- 1Add WhisperAI to ChatGPT and sign in
- 2Ask for the uploader and set language and speakers
- 3Pick the recording
- 4Ask ChatGPT about the transcript
- Find WhisperAI in ChatGPT's app directory, add it, and sign in with your WhisperAI account. The setup walkthrough has the link and the steps.
- Ask ChatGPT to upload a recording to WhisperAI. It opens WhisperAI's uploader in the chat. Before you pick the file, tell it the language, how many people speak, and any names or terms the transcript has to spell right.
- Choose the recording (or up to 20 of them), audio or video, 5 GB or smaller each. It goes straight from your device to WhisperAI. An hour of audio usually comes back in under a minute, depending on the language and the number of speakers.
- Ask for what you came for: a summary, every quote on one topic, the action items, or a translation. ChatGPT reads the transcript from WhisperAI with the speaker labels in it.
Already have the file online? Give ChatGPT a direct link to the audio or video and it can send that to WhisperAI instead, for files up to 200 MB.
A 2-hour transcript is long, and WhisperAI hands it to ChatGPT in parts of about 20,000 characters. Ask about one stretch of the conversation at a time, or ask for an export when you need the whole thing in one file: PDF, DOCX or TXT, or SRT and VTT subtitles with timestamps.
Why isn't ChatGPT transcribing my audio?
It's usually one of five things.
- You're on ChatGPT Free, where audio uploads aren't available.
- You're in Work or Codex, where uploaded-audio transcription isn't available.
- The file is a video with a picture in it. Audio uploads don't take video files.
- The file is bigger than 512 MB.
- The recording runs past 6 minutes. Send it to WhisperAI and let ChatGPT work from the transcript, or cut it into 6-minute pieces.
Can ChatGPT transcribe video?
Not as a video file. Chat accepts MP4 and WebM only when there's no picture in them, so you'd have to pull the audio out first. WhisperAI takes video files as they are, including MP4, MOV and WebM, and transcribes the sound.
Is ChatGPT transcription free?
Not for recordings. Uploading audio and using Record both need a paid ChatGPT plan.
WhisperAI's free plan transcribes 5 minutes a month with no card, speaker labels included, which is enough to see how your own audio comes out. The 2-hour interview needs 120 minutes: Premium ($14.99 a month) includes exactly that, and Business Pro ($24.99 a month) has no minute cap.
Which one should you use?
Check how long the recording runs and how many people talk. A voice memo under 6 minutes? ChatGPT handles it alone. For an interview, a lecture, a meeting with several voices, or anything you want subtitles for, add WhisperAI to ChatGPT and ask it for the transcript. For recordings that run several hours, see how to transcribe long audio files.