Paid plans accept a single recording of up to 10 hours, so a full conference day or a day-long deposition can go in as one file. Long recordings fail differently from short ones though: the engine does not get tired, but drifting microphones, changing rooms and shifting speakers accumulate over hours in a way a ten-minute clip never shows.
Let the browser extract the audio
Video files carry an enormous amount of data that speech recognition never uses. Extracting audio locally before upload turns a multi-gigabyte transfer into a small one, which means fewer interrupted uploads and a faster start.
Set the language and speakers up front
For panel discussions, accented speech or noisy rooms, naming the spoken language explicitly and giving an expected speaker count keeps labels stable across hours of audio. On a clean solo lecture the defaults are usually enough.
Prepare a key-terms list
- Speaker and organisation names as they are actually pronounced.
- Domain vocabulary, acronyms and product names.
- Anything unusual that appears more than twice in the recording.
Split by session, not by size
A conference day transcribes better as separate sessions than as one eight-hour file — not because of a technical limit, but because the resulting transcripts are navigable. Chapters help, but a folder of sessions helps more.
Long jobs continue in the background. Start the transcription, close the tab, and come back to a finished transcript.
Put A Long Recording Through It
Upload a lecture, conference day or deposition, close the tab, and come back to a finished, searchable transcript.
Transcribe A Long File