The day as one text file: an agent reads the speech and proposes the story
The step before any timeline: the day's speech in shooting order, numbered by clip and sentence, read by an agent like a scriptwriter. It proposes story, scenes, hook and cutaways; you argue in text.
Below is one day of footage as text: every sentence spoken, in shooting order, numbered as clip.sentence, with the shooting time and the clip's one-line visual description. Read it like a scriptwriter, not an editor. You are choosing the story before anyone touches a timeline.
Reply in this order:
1. STORY in three sentences: what happened today, to whom, and what changed by the end. If nothing changed, say what the day was about instead.
2. SCENES: 4-8 scenes, each with the sentence range (whole conversations: lead-in, the talk, the reaction), the place, and one line on why it is in the film.
3. HOOK: three candidate lines from the text with a twist (a 'but', a number, a contradiction), each with its sentence number. Say which you would use and why.
4. PROMISES: any sentence about something that will happen later, with the sentence where it happens if it does.
5. CUTAWAYS NEEDED: for each scene, which objects or places the lines mention that must be seen, so the cover shots can be chosen from the same place and time.
6. NOT IN THE FILM: the sentence ranges you would leave out, with one reason each that the person who filmed it could check.
Do not propose timecodes, transitions or music. Text only.
Day:
{paste the day as numbered sentences} In Claude Code, paste the prompt as a message, or run /video-edit which does this end to end.
Steps
- Transcribe every clip with word timings (local whisper is enough; keep a glossary so names stay names), then concatenate in shooting order from container metadata, numbered clip.sentence, with the clip's visual description from the analysis sidecar.
- Run the prompt. Correct the story and the scenes in the reply, as text. This is the cheapest place to change your mind.
- Only then ask for the cut as JSON (see 'write the cut as a script'). The story reply is its brief.
Example
Input line: 07.3 [14:22] (bakery, counter, her face) “If this is the bread they meant, we’re coming back tomorrow.” Output, under PROMISES: 07.3 promise: come back tomorrow. Payoff 23.1 [next day 09:05] “Same bakery. Told you.” Keep both.
Tips
- STT turns unknown names into celebrities. Fix the transcript once with a per-trip glossary before the agent reads it, or the story will star someone famous.
- In Claude Code `/video-edit` runs this step after the sidecar analysis; `/transcribe` produces the word-timed text on the Apple GPU without uploading anything.
| 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 | day-as-one-text-file-story-and-cut-prompt.md |
Related
Write the cut as a script, not a timeline: the JSON storyboard prompt
An agent prompt that writes the edit of real footage as a small JSON file: which sentences, which cover shots, what the cold open is. A complaint becomes a one-line edit, and the render repeats.
Promise and payoff: find the pairs that thread a multi-day film
An agent prompt that reads several days of transcripts and pairs every promise ('tomorrow we'll buy the boat') with the moment it is kept or broken, so the film has a thread, not a sequence of days.
Analyse once: sidecar files beside the sources, shooting time from the container
The storage rule that makes a long film possible a month later: each clip's shots, frame scores, beats and speech live in a file beside it, and shooting time comes from the container metadata.
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.
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.
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.