The critic pass: repeats and sagging scenes first, mid-word joins second
A two-pass review prompt for a subagent or a second session: pass one on structure (repeats, sagging scenes), pass two on joins (mid-word cuts, audio seams). One read finds one kind of problem.
You are the critic for a film cut from real footage. Review it in two separate passes and do not mix them. PASS 1, STRUCTURE (read the cut as text plus the storyboard grid of one frame per shot): - Repeats: the same face, the same place, the same point, more than twice in a row or more than {4} times in the film. Line numbers and frames. - Sagging: any run of more than {40} seconds where nothing new is said, shown or decided. Give the start and end and the one thing that would fix it (cut, or move a scene). - Order: any scene that would land better earlier or later, with why. PASS 2, JOINS (read the word-timed transcript at every cut point): - Lines cut mid-word at a join: the first or last word of a kept sentence shorter than {120} ms. List timecode, the word, and whether the in or out point moves. - Audio seams: a cut where the room tone changes audibly inside a sentence. - Cutaways over moving lips: any cutaway whose window overlaps a word spoken on camera. Output: two numbered lists, PASS 1 and PASS 2, each item with a timecode or line number and a one-line fix. No summary, no praise. If a pass finds nothing, say so for that pass only.
In Claude Code, paste the prompt as a message, or run /video-edit which does this end to end.
Steps
- Give the critic the cut as text, the storyboard grid (one frame per shot: an ffmpeg tile, or Life2Film's scene list exported as frames) and the word-timed transcript. Without word timings pass 2 is guessing.
- Fix pass 1 notes first; they move scenes and invalidate pass 2 timecodes. Re-render, then run pass 2.
- Run the critic in a fresh session or as a subagent. The session that made the cut defends it.
Tips
- Two passes minimum. A single 'review this' returns three structure notes and misses every mid-word cut, because reading and listening are different jobs.
- The storyboard grid catches 'the same face ten times' that every metric passed; include it, not only the text.
| works with | Claude Code: Runs as a skill or a prompt in the terminal; can read your files and write the result. · Claude: Paste into claude.ai or the Claude app. |
| skill | npx skills add fortunto2/solo-factory/video-edit · source |
| cost | Free. No API, no key. |
| status | Unverified: written from the technique, not yet run end to end by us. |
| license | CC BY 4.0 for the text here; skills are MIT. |
| updated | 2026-11-01 |
| for agents | critic-pass-two-passes.md |
Related
Read the film as text before rendering: the last check that costs nothing
The cut, printed as its lines in order, read by someone who never saw the raw footage. Catches the question whose answer was cut, the name said before the person, the 'this' that points at nothing.
A contact sheet of every cutaway with its line under it
The one-page check that catches off-topic cutaways before anyone watches: every cutaway frame in a grid, its sentence under it. Measured once: 12 of 54 automatic cutaways were off topic.
Length follows the material: cut stammers and repeats, not thoughts
The decision rule for how long a film should be. A target length makes you stretch a thin day or chop a full one; the material decides, and the checklist says what may be removed.
Scenes, not best phrases: keep the lead-in, the talk and the reaction
Why a reel of good lines confuses everyone who was not there, and the checklist for keeping a scene whole: the lead-in, the talk, the reaction, with filler trimmed only at the ends.
Take something from every clip, and list the ones you cut with a reason
A rule for cutting someone else's footage, or your own from a year ago: each clip was filmed because it mattered, so each gets a chance, and the rejects get a written reason the person can argue with.
The cutaway rule: what is said, same place and time, never over moving lips
Three constraints for every cutaway in a talking film, and the agent prompt that applies them to a transcript. Kills what viewers read as broken: a picture from elsewhere, a face out of sync.