Postwright Release Note: Long Videos, Edited

My long videos used to go through a hand edit in DaVinci Resolve, where I'd hunt down the false starts, cut the retakes, and trim the pauses where I'd glanced down at my notes. Postwright 0.3.2 does that edit for you. Drop a raw talking-head recording in a folder and it hands back a 1080p cut with cleaned audio and a caption file, then waits for one review before anything goes anywhere.

It edits my own videos now. The rules for what it trims came from reviewing its first edit of one of my recordings: silence over 1.2 seconds shrinks to half a second, a pause while I look down at my notes shrinks to 0.3, and when I say the same few words twice in a row, the first try goes.

What changed

• Long-form auto-edit (free in Studio). It cuts the setup and teardown, false starts, retakes, notes to the editor, filler and dead air, and keeps your last complete take. Your original recording is never moved, renamed or changed.
• A second check before you see it. Another AI pass reads the finished edit for flubs that slipped through and for content cut by mistake, alongside checks for clipped words, clicks at the cuts, loudness and sync. Planning defaults to the strongest model on Claude Code, because in testing a smaller one kept 40 seconds that I'd said, later in the same recording, to cut.
• One review, in the app or on Telegram. You get a reel of every cut with two seconds either side, numbered, so you can put any one back by its number, cut a stretch yourself, or rename the video before you approve it.
• Captions that match the cut. The SRT comes from re-transcribing the finished edit, so its timing lines up with what's on screen.
• Shorts follow. An approved edit goes to the shorts pipeline, every long-form now yields at least three shorts, and you can review those from Telegram as well.
• Shorts sound the same. Shorts get the same audio cleanup, ending at -14 LUFS, with Full, Loudness only or Off in Settings. They also use only a video's first audio track, because an export with separate voice and music tracks could end up with the wrong one.

It needs ffmpeg-full, local transcription, an AI provider and about 15 GB free while it renders. Postwright 0.3.2 is at postwright.app.