YOUR CREATIVE SPACE
Transcribe audio locally
Download the experimental multilingual model, transcribe a short clip and review its text and phrase timing.
—
00:00.00000:10.000
Start with a draft, finish with a review
Optional on-device Whisper base transcription uses a downloadable multilingual model. Choose English or Arabic, create a phrase-timed draft, listen back and correct every cue before applying it.
Worked example
For a short English recording, the model produces text and phrase boundaries that can be applied to the current track, edited and undone. No audio is uploaded to a speech service.
Limits & review
Experimental, not production-grade transcription. Arabic, noise and silence can produce errors. Model/runtime download is about 85 MB; input is limited to 30 MB / 2 minutes. Accuracy and speed depend on content and device.
Method reference · 2026-10-04