Short answer: the recording is the evidence, the transcript is the map, and how good the map must be depends on what you're using it for.
Most transcript use never touches an evidence rule: finding the moment on four hours of video, quoting the report against what the officer actually said, preparing cross, briefing a client. For that work, a fast AI-assisted draft with timestamps is worth more than a slow perfect one, because you check anything that matters against the recording anyway.
When a transcript will be attached to a motion, published to a jury as an aid, or quoted in a way opposing counsel will attack, the standard changes. Courts commonly allow transcripts as listening aids when a sponsoring witness attests they're accurate; that attestation needs a human who actually checked every word against the audio.
That's exactly the difference between our two tiers: the machine draft for working the case, and the human-verified transcript, checked word for word against your discovery, for the moments that count. Neither is a certified court-reporter transcript, and we label them accordingly.
Work fast from drafts, verify before you rely on anything, and when a line is dispositive, quote the recording rather than the transcript. Keep the speaker labels grounded in the discovery too, which is why we ask for the reports and every camera angle, not just the clip.
First human-verified case is free for new clients.
See the live sample