Speak AI transcribes 93 languages and regional variants across 65 base languages, including 13 Arabic dialects, with automatic language detection. Finished transcripts translate into 111 languages .
- 50 languages support live, real-time transcription
- 31 support custom vocabulary for names and jargon
- 28 translation languages support dubbed audio
The full list
The complete per-language table (transcription, live, vocabulary, translation, dubbing) is
generated from the same source file the product reads, so it can never drift. Rendered at build
time from data/facts.json.
Working across languages
- Auto-detection picks the language per file. Short or noisy files can detect wrong; set the language at upload when you know it. Fix a wrong-language transcript
- Mixed-language audio: set the dominant language rather than relying on detection, then correct in the editor.
- Translation happens after transcription, on the finished transcript: Translation.
Need a language you don’t see, or evaluating multilingual coverage for a team? Book a demo and bring a recording in that language.
Related: Transcription · Translation · Custom vocabulary