Languages
Transcription in 33 languages
Scriptorium transcribes speech in 33 languages, plus automatic language detection, and you can try a file free without signing up (up to 15 minutes). Pick your language for notes on its dialects, script and quality; the groups below say honestly where the small model has more or less to work with and where it needs more editing.
Grouping by how much speech data Whisper-family models are known to have seen per language. It is our own rough guide, not a measured score, and real results depend mostly on audio quality. Not sure which language it is? Audio to text detects it for you.
Large amount of training data: still a draft
Good amount of training data: expect corrections
Less training data: expect more editing
Least training data: treat the draft as a rough guide
Read this site in your language
These pages are written natively, not machine-translated word for word.
More reading: how to transcribe non-English audio, recordings with more than one language. All use cases are on the use cases page.
Questions
How many languages can it transcribe?
Are all languages equally accurate?
Which language setting should I use?
Can it translate?
Is it free?
Try it with your own file
Upload a recording and read the transcript.