transcribe.free
Free audio and video to text
8 followers
Free audio and video to text
8 followers
Turn audio or video into accurate text in minutes. Free, no account needed — 3 files a day, up to 30 minutes each. Export TXT, SRT, VTT. Files deleted after 24 hours.
How well does it handle background noise or overlapping speakers, and is there any limit on supported languages beyond just transcription accuracy?
@havvaq6ab Whisper-large-v3 degrades gracefully on noise, muffled or windy recordings come out with more errors but still readable, and only fully masked speech is lost. Overlapping speech is the true weak spot: the model transcribes one dominant voice, and diarization (paid) only labels who spoke each turn, it can't untangle people talking at once. And regarding language limits beyond accuracy: Yes, Whisper has a hard set of ~99 languages, and audio outside it produces confident garbage rather than an error; mixed-language speech also gets forced into one language per segment. Our own dropdown additionally only pins 19 languages, so the other ~80 are reachable via auto-detect only.