Questions, answered.

Whisper-class accuracy - typically 95%+ on clear audio. Accents, crosstalk, and heavy background noise lower that; a good mic raises it. Every word is timestamped so errors are fast to find and fix.

No minute caps, no file counts, no throttling after some hidden threshold. The only limits on the Unlimited plan are 10 hours per single file and a fair-use ceiling of 24 hours of audio per day - more than a day contains, so no human ever hits it. Fair-use applies to automated bulk abuse, not humans with lots of tape.

MP3, WAV, M4A, FLAC, OGG for audio; MP4, MOV, WEBM, MKV for video. We pull the audio track from video automatically - no conversion needed on your side.

90+ languages, auto-detected - including English, Uzbek, Russian, Spanish, Mandarin, Arabic. Mixed-language recordings work; each segment is tagged with its detected language.

About a minute per hour of audio on the priority queue. Free-plan files queue behind paid ones at busy times but still typically finish within a few minutes.

Files are encrypted in transit and at rest, processed automatically, and never used to train models. Delete a transcript and the source file goes with it, immediately.

Yes - one click in Settings, no email required. Yearly plans are refunded pro-rata within the first 30 days. Your transcripts stay downloadable on the free plan after you cancel.