Can ChatGPT transcribe audio? What works, what doesn't, and when to use something else
Short answer
Yes, on paid plans: ChatGPT accepts audio files up to 512 MB, and the Free plan doesn't support audio uploads as of October 2026. It works for short, clear recordings if you ask for a verbatim transcript, but long files may time out and speaker labels and timestamps aren't reliable.
Short answer: yes, on paid plans, and it works well for short, clear recordings. ChatGPT's help center says paid plans accept audio uploads (MP3, M4A, WAV, FLAC, OGG and others, up to 512 MB). The Free plan doesn't support audio uploads as of October 2026.
For a five-minute voice memo, that's often all you need. For a one-hour meeting, an interview with three people or anything you'll need to quote later, a general-purpose assistant has real limits. Here's how to do it and where it falls short.
How to transcribe an audio file with ChatGPT
- 1
Check your plan and file
You need a paid plan and an audio file (mp3, m4a, wav). According to ChatGPT's help center, video files such as an mp4 from your phone aren't accepted by audio uploads, so extract the audio first.
- 2
Upload it and ask for a verbatim transcript
Attach the file with the + button and be explicit: "Transcribe this word for word. Don't summarize. Start a new line when the speaker changes." Without that, a general assistant may give you a summary instead.
- 3
Check the result against the audio
Spot-check a few passages, especially names and numbers. Processing is best-effort: very long recordings may time out, so split them into shorter parts if that happens.
Other ways ChatGPT handles voice
- Dictation: the microphone in the message box turns what you say into a typed prompt. It's for your own voice, not for a file.
- Record mode: on the macOS desktop app, Plus, Pro, Business, Enterprise and Edu users can record a meeting or voice note (up to 4 hours per session) and get notes in a canvas. It's Mac-only and produces notes, not a speaker-labeled transcript.
- Other assistants, such as Google's Gemini, also accept audio on some plans and devices, with length limits that depend on the plan.
Where a general assistant falls short
A chatbot answers your request; it doesn't guarantee a complete, word-for-word record. That's fine for a quick idea, risky for an interview you'll quote or meeting minutes people will rely on.
| Need | ChatGPT | Dedicated transcription tool |
|---|---|---|
| Long files (1–4 hours) | Best-effort; may time out | Built for it |
| Who said what | Only if you ask, and not reliable | Speakers separated automatically |
| Timestamps | Not reliable | On every line |
| Export | Copy and paste | TXT, PDF or Word |
| Free option | No audio uploads on Free | Usually a free monthly allowance |
Privacy: will your audio be used for training?
On personal ChatGPT plans, your conversations, including uploaded files, may be used to improve the models unless you turn off "Improve the model for everyone" under Settings > Data Controls (or use a Temporary Chat). Business, Enterprise and Edu workspaces aren't used for training by default. If the recording includes other people, check this setting and ask for their consent before uploading.
When Dolcenota is the better tool
Dolcenota is built only for recordings. You upload audio or video (only the audio is transcribed), up to 4 hours and 2 GB per file, or record in the browser. In a few minutes you get a transcript with speakers separated (Speaker 1, Speaker 2, until you rename them), a timestamp on every line, a summary with action items by person, and answers to your questions grounded in the transcript. It works in English and Spanish, including recordings that mix both, and exports to TXT, PDF or Word.
Your recordings are processed on Amazon Web Services (AWS), stored encrypted and never used to train AI models. There's a free plan with 60 minutes a month and every feature.
Frequently asked questions
Can the free version of ChatGPT transcribe audio?
Not from an uploaded file as of October 2026: audio uploads are for paid plans. You can still dictate with the microphone.
What's the file size limit for audio in ChatGPT?
ChatGPT's help center lists audio files up to 512 MB, but long recordings may time out because processing is best-effort.
Can ChatGPT tell speakers apart?
You can ask it to, but it isn't a dedicated feature, so labels can be wrong or missing. Transcription tools separate speakers automatically.
Can I upload a video to ChatGPT to get a transcript?
Video files aren't accepted as audio uploads, so extract the audio first. Dolcenota accepts mp4, mov and webm and transcribes the audio track.
Does Dolcenota use my recordings to train AI?
No. Recordings are processed on AWS, stored encrypted and never used to train AI models, and you can delete them anytime.
Need the full transcript, not a summary?
Upload your recording and get every word, separated by speaker with timestamps, plus a summary and answers. Free to start.
Start free