01The problem
Tools that edit podcast video automatically assume a studio: a microphone and a camera for every person, so the software can tell who is talking by checking which track is loud. Most shows do not have that. A camera on a table, a phone or a Zoom call gives you one picture and one mixed audio track, and someone ends up cutting it by hand.
02What I built
Cutroom works out who is talking from the recording itself: whose voice it is, and whose lips are moving. You drop in a one-camera recording of two, three or four people. It writes the transcript, finds each person's face, matches voices to faces by lip movement, and cuts in close on whoever is speaking. Then it hands you every decision to change.
- A transcript with a timing for every word, and speaker detection on a single mixed track, including when people talk over each other
- An editor with a live preview that matches the export, and a framing timeline you can zoom, drag and undo
- Dead air and filler words found and shown on the timeline, so you can put any cut back
- Export to MP4 at the source resolution, with optional captions, a subtitle file and an MP3
Everything runs on your own computer. The recording is never uploaded, and there is no account or subscription.
03Where it stands
It is a developer preview. The whole path from recording to edited video works and has been run on real multi-person episodes, but it has rough edges, and the Mac and Windows installers are not signed yet.
04Results
A 53-minute, four-person episode, cut in 15 minutes.