Can ChatGPT transcribe audio?
Yes, ChatGPT can transcribe audio – on a paid plan, and for short recordings. You attach a file to a chat, ask for a transcript, and the text comes back in the reply. As of November 2026, OpenAI’s own help page puts the limit for an uploaded recording at six minutes in the chat model that does the transcribing, and says audio uploads are not available on the free plan.
So the honest answer depends on two things: how long the recording is, and whether it is all right for it to leave your phone. A two-minute voice note of your own, on a plan you already pay for, is exactly what ChatGPT is good at – and you can ask for a summary in the same breath. An hour-long lecture, a recording of someone else’s voice, or a phone with no signal is where an app that transcribes on the device does the job instead.
What ChatGPT does with a recording
On a paid plan you attach the audio – MP3, M4A, WAV, OGG, FLAC and a few other formats – and ask for a transcript. The file is uploaded to OpenAI and transcribed there. A recording longer than the limit may be processed in smaller sections when ChatGPT’s data analysis tools are available, but OpenAI calls that best effort and warns that very long recordings may time out.
OpenAI is also candid about quality: transcripts may contain errors, telling speakers apart may be unreliable, and results vary between languages. Which ChatGPT apps accept audio attachments depends on the app version, and the help page does not list them.
On a Mac there is a second route. Record is a mode of the ChatGPT desktop app on paid plans: it records a meeting, transcribes it and writes notes. It lives on the Mac only – not on the iPhone and not in a browser.
Where the audio goes
Everything above happens on OpenAI’s servers. An uploaded file can be saved in your ChatGPT library separately from the conversation, so deleting the chat does not remove it, and whether your content is used to improve models depends on your plan and your data settings. None of that is unusual for a cloud service, and for a note to yourself it hardly matters.
It matters when the voice is not yours. A patient, a source, an interviewee, a colleague in a meeting – where their recording goes to become text is a decision you are making about them. Kaseta transcribes on the iPhone, in 99 languages, with no account and nothing uploaded; once its models are downloaded, it works in airplane mode.
When ChatGPT is the better tool
- You already pay for it and the recording is short. Attaching a two-minute note is quicker than installing anything.
- You want more than a transcript. A summary, a list of decisions, an email drafted from what was said – ChatGPT does that from the same upload. Kaseta transcribes; it does not write summaries.
- You want to ask questions about the recording. “What did they say about the deadline?” is a conversation, and a transcription app does not hold one.
- Your meetings happen on a Mac. Record captures the call where it happens, which a phone lying on the desk cannot.
When an app on the phone is the better tool
- The recording is long. A lecture, an interview or a meeting runs well past the chat limit. Kaseta transcribes the whole recording on Pro; on the free plan you get the text of the first 5 minutes of each one, and the audio is always kept in full.
- You are on ChatGPT’s free plan. There is no audio upload there at all.
- The audio should not leave the phone. Someone else’s voice, a confidential meeting, a workplace that does not allow cloud tools.
- There is no signal. A plane, a lecture hall in a basement, a building with bad reception.
- You need a file, not a reply. Kaseta exports Plain Text, Markdown, SRT, WebVTT and JSON with timecodes on every plan, and labels speakers when you ask it to.
Use both
The two fit together better than they compete. Transcribe the recording in Kaseta, copy the text, and paste it into ChatGPT with your request – “summarise this”, “list the action items”, “turn it into an email”. The six-minute ceiling is on audio, not on text, and you decide which parts of the transcript leave the phone. The recording itself never does.
How to get a memo into Kaseta is in the voice memo guide, why on-device recognition needs no connection is on the offline transcription page, and long conversations with several voices are covered in interview transcription.