Can ChatGPT transcribe a lecture recording?

Short answer first: only on a Mac with a paid plan. ChatGPT Record, in the macOS desktop app on Plus (20 US dollars a month), Pro, Business, Enterprise and Edu, records for up to four hours per session and returns a transcript with editable notes. A recording you already have is a different story: OpenAI's upload help lists documents, spreadsheets and presentations, not audio. Developers can use the API, which transcribes files of up to 25 MB for an estimated 0.0045 US dollars a minute. Otter, Turbo AI and Notibo handle the recording and the notes in one app.

We read OpenAI's help centre, its API documentation, and the pricing and privacy pages of the other vendors on 25 September 2026. If a company does not publish a number, this article says "not published" and does not estimate one.

What does a lecture transcription tool have to handle?

The word transcribe hides more than one job: getting the audio in, turning all of it into text, and deciding what happens to the recording afterwards. ChatGPT handles each of these differently depending on the app, the plan and whether you are a developer, so a yes or no without those details is not much use.

Does it take the file you already have?

Many lectures reach a student as a file: a phone recording, an M4A from a voice memo app, an MP4 the department posted. A tool that only records live is no help with those, and a tool that only takes uploads is no help in the lecture hall. Decide which of the two you need before you compare anything else.

Will a 90-minute lecture go through in one piece?

Caps hide in two places, minutes per session and megabytes per file. By our own arithmetic, a 90-minute lecture recorded as an MP3 at 128 kilobits a second weighs roughly 86 MB, more than three times the 25 MB OpenAI accepts per file in its transcription API. Anything over a cap has to be cut into pieces, and the text stitched back together afterwards.

Does it say what happens to the audio afterwards?

A lecture holds the lecturer's voice and often other students' questions. You want a plain statement of whether the audio is kept, for how long, and whether it trains anyone's model. OpenAI answers these questions in detail, but the answer differs between Record, Voice, Dictation and the API, so read the one that applies to you.

How does ChatGPT compare with tools built for transcription?

The table sets the two OpenAI routes against four tools made for recordings. Prices are in the currency each vendor's page showed; Notibo's is in euro.

The two OpenAI rows are separate products with separate rules. Record belongs to your ChatGPT subscription. The API is a developer service with its own bill, its own file limit and its own data terms. Mixing them up leads to wrong answers about limits, price and training.

Can you upload an audio file to a ChatGPT chat?

This is the question when the lecture is already on your phone, and OpenAI's own pages do not describe it.

• Audio is not on the supported list. OpenAI's File Uploads FAQ names text files, spreadsheets, presentations and documents, with examples such as summarising a research paper or comparing two documents.
• The hard limit is 512 MB per file. Documents are also capped at 2 million tokens and images at 20 MB, and Free accounts can upload three files a day.
• Files live as long as the chat. Delete the chat and its files leave OpenAI's systems within 30 days, unless OpenAI has to keep them for security or legal reasons.
• No help page describes transcribing an upload. We searched the help centre and the ChatGPT release notes and found transcription described for Record, Voice and dictation, not for uploaded files.

If an MP3 goes through in your chat anyway, you are relying on behaviour OpenAI does not describe, with no stated limit on length, no language list and no promise it will work the same way next month. For a lecture you need in full, that is a weak base. If what you want is to question a recording rather than read it, NotebookLM takes audio uploads, as our NotebookLM guide explains.

What do Record, Voice and Dictation actually do?

ChatGPT has three features that listen, and only one of them is meant for a lecture.

• Record transcribes and summarises up to four hours. OpenAI's Record page limits it to the macOS desktop app on Plus, Pro, Business, Enterprise and Edu, caps a session at 240 minutes and saves the notes as a canvas you can edit or rewrite.
• Record is strongest in English. OpenAI says accuracy in other languages "is improving but can vary", and when it cannot name a speaker it uses labels such as Speaker 1.
• Voice is a conversation. The ChatGPT Voice page describes speaking with ChatGPT and hearing a reply, and says Voice transcripts are not verbatim records.
• Dictation turns one spoken prompt into text. OpenAI's dictation FAQ describes recording a message with the microphone icon and editing the text before you send it.

On Windows, on a phone, and on the Free and Go plans there is no Record button, so ChatGPT cannot capture a lecture for you there. On a Mac with Plus, Record is a real option for an English lecture at no extra cost; the two Pro tiers cost 100 and 200 US dollars a month, and new sign-ups to the 200 tier have been paused since 10 September 2026. OpenAI asks you to get consent before recording others, and Enterprise and Edu workspaces keep Record off until an owner enables it.

What does OpenAI do with your recordings?

The rules depend on the account and the feature, and for personal accounts the default leans towards training.

• Personal plans can train OpenAI's models unless you opt out. OpenAI's data page says that for services for individuals, such as ChatGPT, "we may use your content to train our models". The switch is Improve the model for everyone under Settings, Data controls.
• Record deletes the audio but may keep the text for training. The Record page says audio is used only for transcription and then deleted, while transcripts and canvases from Free, Plus and Pro may be used if that setting is on.
• Feedback reaches past the opt-out. A thumbs up or down can bring the whole conversation into training even after you have switched training off.
• Business, Enterprise, Edu and the API are excluded by default. The same data page says OpenAI does not use their inputs or outputs to improve its models by default.

With ChatGPT Edu from your university, the training question is settled for you, although Record still needs an owner to enable it. On a personal Plus account, switch the setting off before you record, and delete the chat when you are done: OpenAI says the transcript and canvas then leave its systems within 30 days.

Is the OpenAI API a cheaper way to transcribe?

The developer platform is a separate product with its own bill, and OpenAI's Plus page says so plainly: API usage is not part of a ChatGPT subscription.

• 25 MB per file, in seven formats. OpenAI's file transcription guide accepts mp3, mp4, mpeg, mpga, m4a, wav and webm, and tells you to split longer recordings into chunks of 25 MB or less.
• About 0.0045 US dollars a minute. The API pricing page lists that estimated cost for gpt-transcribe, the model the guide recommends, which works out to roughly 0.41 US dollars for 90 minutes of lecture.
• Languages by code. gpt-transcribe takes the language codes you expect in the audio, and the guide says the older whisper-1 supports 98 languages, with accuracy that varies between them.
• Kept out of training by default. The data page puts the API with the business products.

On price the API is hard to beat, but it returns text and nothing else, and whether bare text is what you need is its own question, covered in transcription versus notes. It also needs an API key, some code and a way to cut the file. A student who already writes Python can adapt the example in the guide. For everyone else, an app does the same job behind a button and adds the notes on top.

Is Otter a closer match for a lecture?

Otter records in the room on a phone, a laptop or the web, and shows the transcript while it forms. That is the part ChatGPT offers only on a Mac.

• Free: 300 minutes a month, 30 per conversation. Otter's pricing page also limits Basic to three file imports in total.
• Pro: 16.99 US dollars a month, 90 minutes per meeting. Billed yearly it comes to 8.33 a month, and students with a .edu address get 20 per cent off.
• Six languages. Otter transcribes English, Spanish, French, German, Japanese and Chinese.
• AWS in the US, training on de-identified data. Otter's privacy policy names Amazon Web Services in the United States for storage, and says Otter trains its own AI on de-identified audio recordings and on transcriptions.

Next to ChatGPT Record, Otter runs on more devices but for less time per session: four hours on a Mac against 90 minutes on Otter Pro. Otter summarises, but its pricing page lists no flashcards, so revision stays a separate step. Our Otter alternatives guide goes further into its limits.

Where do Turbo AI and Whisper apps fit?

Two more routes suit different students: an app that turns the recording into study material, and an app that only transcribes but keeps the audio on your own device.

• Turbo AI makes notes, quizzes and flashcards from a recording. Its App Store listing describes recording a lecture in the app, and a Turbo AI blog post from July 2026 prices it at 20 US dollars a month, the same as ChatGPT Plus.
• Turbo AI runs on other companies' models, OpenAI's among them. Its privacy policy says it operates no in-house AI models and sends content to OpenAI, Grok, Anthropic and Modal Labs.
• Aiko runs OpenAI's Whisper model on your own device. Aiko's listing on the US App Store sells it for 24 US dollars, lists 100 languages and requires iOS 26 or macOS 26.
• Aiko returns text, not notes. The listing says it does not transcribe live while recording, has no speaker detection and offers no editing inside the app.

For a student who wants the studying done, Turbo AI is closer than ChatGPT to what lecture notes usually means, though its free tier and caps are unpublished. For a recording that should stay on your laptop, Aiko is the Whisper-based app we could verify, and our guide to the best transcription apps for students covers more of them.

Which should you use for a lecture recording?

• If you have a Mac and already pay for ChatGPT Plus, try Record on your next English lecture. It runs up to four hours at no extra cost; switch off model training first.
• If the lecture is already a file, skip the chat window. Use the API if you can write a short script, or an app that accepts uploads.
• If you record on a phone or on Windows, ChatGPT has no Record button for you. Otter covers English and five other languages, with 30 minutes per conversation on the free plan.
• If the audio must stay on your device, Aiko runs Whisper locally for a one-time 24 US dollars.
• If you want flashcards from the same recording, choose a tool that makes them, because Record returns a transcript and notes, not a deck with a review schedule.

Test one real lecture from start to finish before you pay for a year of anything.

Where Notibo fits

Notibo runs in the browser on a computer, which matters if your laptop runs Windows. Once the lecturer has agreed, it can record the microphone; Chrome and Edge also let it capture the audio playing in a browser tab, and Windows lets it capture everything the computer plays. A single session can last two hours. For a lecture you already have, it accepts MP3, M4A, AAC, WAV, OGG, Opus, FLAC, WebM, MP4 and MOV files of up to 50 MB.

Check that 50 MB limit against your files. By the arithmetic above, a 90-minute MP3 at 128 kilobits a second is too big, so either record in Notibo itself or export the file at a lower bitrate. The transcript is written in whatever language was spoken, detected automatically among 89, and structured notes and a deck of flashcards follow, with reviews timed by spaced repetition. The transcript and notes can be edited and downloaded as a PDF or Word file.

Your files sit on Supabase in Frankfurt, inside the EU, and the transcription and note writing are done by processors in the US under standard contractual clauses. The recording itself is removed from storage as soon as the transcript and notes are saved. Notibo does not train on your content, and every transcription request opts out of model training. Pro is free for 14 days, with 240 minutes of audio, and needs no card. After the trial the free tier covers 30 minutes of audio a month. On Pro, 1,200 minutes a month cost 9.99 euro a month or 88.99 euro a year. The transcription page has the details.