Transcribe audio free, lectures to study notes
Record the lecture, upload it, and get an accurate transcript you can turn into study notes and search later. Only need the free tier and want the real limits first? We compared the best free transcription tools, minute caps and all.
Last updated: August 2026
Around 100 languages, no account, 10 recordings a day, no watermark.
No account. 10 recordings a day. Files up to 60 minutes here, and it resets at midnight.
Plain transcript and subtitles right here. Bleeping, translation and burned-in subtitles live in the full tool.
Study from text, not a 2-hour recording
Re-listening to a whole lecture to find one point is slow. A transcript lets you skim, search for a term, and turn the key parts into notes and flashcards. It's also a big help if the lecturer spoke fast or in your second language.
- Skim and search instead of scrubbing audio
- Turn key sections into notes and flashcards
- Catch what you missed if the lecturer spoke quickly
- Works for lectures in about 100 languages
How to transcribe a lecture
- Record the lecture on your phone or laptop (with permission).
- Save it as MP3, M4A or a video file.
- Upload it to ScribeGrab (up to 60 minutes / 1.5 GB).
- Download the transcript as TXT and build your notes from it.
How well a lecture comes out
A clear recording from near the front transcribes well; a faint recording from the back row is harder for any tool. Technical terms are worth a quick proofread. See audio to text for the general flow and how accurate it is. Always record with permission. Using an AI to transcribe audio from a lecture hall works best when the recorder sits near the speaker, and the SRT keeps a timestamp on every line so you can jump back to the moment in the recording.
From a two-hour recording to notes you'll actually revise from
A transcript is not notes. It's a 15,000-word document that contains your notes somewhere inside it, and if you stop at the transcript you've moved the problem rather than solved it. The step that makes the difference takes about twenty minutes per lecture and turns reading into remembering.
Work through the text once and pull out only three things: the claims (what is being asserted), the definitions (any term the lecturer bothered to define is a term that appears on an exam), and the examples the lecturer chose, because those are what you'll be asked to apply. Everything else, the admin, the digressions, the repeated explanations, can go. Rewriting a claim in your own words is the part that does the learning; copying it does almost nothing.
The timestamps are what make this fast rather than tedious. Download the SRT alongside the plain text and every line carries the moment it was said, so when a passage makes no sense on the page you can go straight to it in the recording instead of hunting. The click-to-play transcript on the results page does the same thing without leaving the browser: click a line, hear that moment. Use it on the two or three passages per lecture where the lecturer said something that clearly mattered and you didn't follow it, since that's where re-listening actually pays.
One more use worth knowing: a text transcript is searchable across an entire term. Keep them in one folder and searching for a concept will tell you every lecture that touched it, which is the single most useful thing you can have a week before an exam.
Recording a lecture with the phone you already have
Accuracy is decided before you upload anything, and lecture halls are genuinely hostile: a distant speaker, hard parallel surfaces, ventilation noise and 200 people shifting in their seats. The good news is that the fixes are free.
Distance beats everything. Every metre closer to the lecturer helps more than any setting on your phone. Sitting in the first few rows will do more for your transcript than any amount of processing afterwards. Point the phone at the speaker, not at your notebook, and leave it flat on the desk rather than in a bag or pocket, since fabric muffles speech far more than it muffles noise. If your device has a "voice memo" or mono speech mode, use it: it's tuned for exactly this.
Read more
Watch the length. ScribeGrab handles files up to 60 minutes each, which covers most single lectures but not a three-hour block or a full seminar day. If your recording is longer, split it in a free audio editor and upload the halves separately; you still get 10 recordings a day, so two uploads cost nothing extra. Splitting on a natural break also gives you two shorter documents, which are easier to work through than one enormous one.
Expect trouble in predictable places. Technical vocabulary and names are the main casualties, especially in maths, medicine and law: fix them once with find-and-replace and the rest reads cleanly. Questions shouted from the back of the hall often come out garbled, and there's no fix for that beyond noting the timestamp. If the lecturer speaks a language other than the slides, transcribe first and translate second; language detection is automatic across around 100 languages, but it works on what was actually spoken.
Finally, the boring but important part: check your institution's rules before you record. Most universities allow recording lectures for personal study and many publish their own recordings anyway, but permission is usually tied to personal use, and sharing or uploading a recording is a different question from making one.
Got a photo of a handout instead of an audio recording? ScanReviver's image to text reads the words straight off the image with OCR.
If the handout is already a PDF and you want lecture notes as lightweight markup instead of a flat text dump, ScanReviver's PDF to Markdown keeps the headings and lists intact.
By Sam Ridder, I build and run ScribeGrab on my own hardware, on my own. Who I am.
Frequently asked
Is it free for long lectures?
Yes. Up to 60 minutes and 1.5 GB per file, with 10 recordings a day and no watermark or account. You get a free transcript of the entire lecture, the full transcript and not a preview.
What if the lecture is longer than an hour?
A two-hour lecture and a two-minute memo go through the same upload box, but free files go up to 60 minutes. Split a longer recording in two, or use the paid pass, which takes files up to 4 hours. Processing takes roughly two minutes per hour of audio, so it's back well before your next class.
Will it handle technical terms and accents?
Mostly, but proofread specialised vocabulary and names, which are hard for every transcriber. The lecture transcription uses a general speech model, not one trained on academic vocabulary. A strong accent or an echoey hall makes it a little harder; sitting closer to the speaker helps more than any setting.
What if the lecture isn't in English?
About 100 languages are auto-detected, so most lectures work without changing any settings.
Can I record and upload later?
Yes. Record with permission, then upload the file whenever you like; you get 10 recordings a day, reset at midnight. You can transcribe lectures from the same course one after another.
Do I need the video, or is the audio enough?
Either works for transcribing. Upload whatever you have to transcribe to text: audio-only files (MP3, M4A, WAV, FLAC, OGG, Opus, AAC) and video files transcribe the same way. A recorded presentation works like a lecture: only the spoken parts become text, so a slideshow with no narration produces no transcript for that stretch.
Does it organise my notes or add punctuation?
Punctuation is added automatically and the TXT has one sentence per line, but it's a plain transcript. It isn't an AI note taker: turning it into organised study notes is the step described above, and it's yours to do.
Do I get timestamps?
Yes. SRT and VTT with timestamps come free alongside the plain transcript, so the free option of this AI transcriber includes what paid tools often charge for.
Does it separate speakers in a seminar?
Yes, up to eight distinct voices get their own speaker label automatically. With only the lecturer talking, no labels are added.
How to transcribe a voice memo I recorded after the lecture?
The same way as the lecture itself: upload the memo and the text comes back in a minute. Notes you dictated to yourself, in a voice memo or any recording app, transcribe like any other file; length doesn't change the steps.
Related: Obsidian voice notes.
Can I transcribe online from a phone or tablet?
Yes. Any device with a browser works the same way: it's a transcribe website, not an app, with no account or email needed. Upload the phone recording directly, no conversion first. The AI voice recognition handles a lapel mic, a laptop or a phone the same way.
Is my recording kept private?
Yes. The file is deleted right after processing, with results wiped within 45 minutes (24 hours with any pass, free or paid).
How does this compare to a human transcription service?
It's free, where a human service typically charges per minute of audio. Among AI transcription tools the trade-off is the same as with any transcribe AI tool: you let the AI transcribe fast, and you check names and terms. A human transcriber is worth paying for when every word must be certified.
Can I transcribe audio file free from FLAC instead of MP3?
Yes. To transcribe text from speech it doesn't need MP3: FLAC, OGG, Opus and AAC all upload directly alongside MP3, WAV and M4A, and you can transcribe audio free in any of them.
Which AI transcribing tools work for a lecture recording?
Any tool that takes a saved audio or video file can do it. ScribeGrab works from the recording after the class, not live, and gives you a TXT transcript plus SRT and VTT with timestamps. What to compare is the limits: here it is 10 recordings a day, up to 60 minutes each, with no account.
Does an AI transcribing tool need the whole lecture in one file?
No. You can upload the first part and the second part separately. A lecture longer than 60 minutes has to be split anyway, and the transcripts are easy to join afterwards.
Is there a note transcriber that turns a lecture into finished notes?
It transcribes, it does not summarize. You get the full text, and the notes come from you picking the claims, definitions and examples out of it, as described above. That takes about twenty minutes per lecture.
More guides: transcribe a podcast · MP3 to text · transcribe an interview · audio to text