How to transcribe a long recording
A recording longer than two hours or bigger than 64 MB has to be split before it can be transcribed — the service rejects bigger files instead of failing halfway. Cut a long file into parts of up to 120 minutes, transcribe each part, and Pro ($10 a month paid yearly) covers about 12 such two-hour parts a month inside its 1,500-minute allowance.
The limits, and what happens when you cross them
| Limit | Value | What happens when you cross it |
|---|---|---|
| Size of one file | 64 MB | the upload is rejected with a clear error |
| Length of one file | 120 minutes (two hours) | the upload is rejected with a clear error |
| Free plan | 3 files a day, up to 30 minutes each | the next upload is refused until the next day |
| Pro fair use | 1,500 minutes of audio per calendar month | a clear limit error, the account stays open |
| Team | committed volume, written down | agreed case by case |
| API | metered by volume, no subscription | billed on the audio you send |
Step 1 — split the file
Any tool that cuts an audio file works. With ffmpeg it is one line, and -c copy means no re-encoding, so it is fast and lossless:
ffmpeg -i lecture.mp3 -f segment -segment_time 120 -c copy part-%02d.mp3The result is a set of parts, each at most 120 minutes long. If a part is still over 64 MB, shorten it or lower the bitrate — 64 MB is about 70 minutes of audio at 128 kbps (calculated).
Step 2 — upload the parts
Each part is a separate file, so each part is a separate transcript. The web app has drag and drop for that; a pipeline should use the chunked upload API, which takes a large file in 8 MB pieces instead of one request:
| Step | Call | What it does |
|---|---|---|
| 1 | POST /v1/uploads/init |
opens an upload and returns its id |
| 2 | PUT /v1/uploads/{id}/chunk?index=N |
sends one 8 MB piece of the file |
| 3 | POST /v1/uploads/{id}/complete |
assembles the pieces into a file and queues the transcription |
Step 3 — put the parts back together
Here is the honest part: the app keeps every part as its own transcript, and the SRT or VTT export of a part carries that part's own timecodes. To make one continuous file you have to add the offset of each part to its timecodes. The API gives you the numbers to do that — every segment comes back with start_s and end_s — so a short script can shift them and concatenate the files.
How long a long recording takes
Our engine cuts audio into 22-second fragments with a 3-second overlap and assembles the timecodes afterwards, so a two-hour file is roughly 330 fragments. How long that takes depends on the load on the service, and we do not publish a fixed speed: the widget shows the status while the job runs.
| Recording | Minutes | Parts a month inside 1,500 minutes (calculated) |
|---|---|---|
| one two-hour lecture | 120 | 12 |
| one 90-minute seminar | 90 | 16 |
| one 45-minute recording | 45 | 33 |
Put a file on the page and read the first lines.
The widget below is the same one as on the home page: file → transcript with timecodes and speakers → export. The transcript is saved to your account.
Drop an audio or video file
or — mp3, wav, m4a, mp4, up to 64 MB and two hours
Transcription runs under your account: sign in or create a free account.
No account yet? The free plan needs no card. Create a free account or sign in.
Questions this page answers.
What is the longest file I can transcribe?
120 minutes and 64 MB per file, on every plan. A file over either limit is rejected at upload with a clear error, before any work is done.
Why was my file rejected?
Either it was over 64 MB or over 120 minutes, or — on the free plan — over 30 minutes. The error message says which one, and the widget shows it as well.
How do I split a file without ffmpeg?
Any audio or video editor that can cut and export will do: cut at a pause, export each part as mp3 or wav, and upload the parts one by one. Keep every part under 120 minutes and 64 MB.
Can I upload a 2 GB recording?
Not as one file. Use the chunked upload API — the pieces are sent in 8 MB requests — but the assembled file still has to fit the 64 MB and 120-minute limits, so split the recording first.
How long does a two-hour file take to transcribe?
The engine works through 22-second fragments with a 3-second overlap and then assembles the timecodes. The total time depends on the load on the service, so we do not promise a fixed number; the widget shows the status while it runs.
Do the timecodes of the parts line up?
Each part is transcribed on its own, so each part starts at 00:00. The API returns the start and end of every segment, so you can add the offset of each part to the timecodes and concatenate the exports.
Related pages.
When the recording is a call.
A long video usually needs the same split.
The general guide.
The fair-use allowance of 1,500 minutes a month.
Mark Burmistrov, Founder.
Every number on this page comes from the service contract, the pricing page or our own measurements. The author of the page and the publisher of the site are the same person. Questions and corrections: support@transcrib.si.