Generated b-roll or real: the decision rule
When a generated cutaway is fine (faceless, no logos, no claim about a real place or time) and when only footage from the same day will do. Five questions, in order, before you open a video model.
GENERATED OR REAL, answer in order; the first REAL wins
1. Is the line about something that happened to the speaker, here, today? REAL. A film about a day uses the day's pictures; a generated beach over 'we swam at dawn' is a lie the viewer can feel even when they cannot name it.
2. Does the shot need a face, a known person, or a recognisable real place? REAL. Generated faces drift between clips and generated landmarks are wrong in ways locals see at once.
3. Would the shot carry a logo, a product label, a screen with real text, or anything a viewer might read? REAL, or a screen recording. Models still garble text and invent brands.
4. Is the line general (how a thing works, what a category of object looks like, an abstract process) and would the shot be faceless with no logos? GENERATED is fine. Prompt it literally: 5 s, one action, 'no on-screen text, no people's faces', in the film's style lock.
5. Do you have the real thing but shot badly (shaky, dark, 2 seconds)? REAL, two seconds of it. Bad real footage of the right thing beats a clean generated clip of a similar thing, in a film about real events.
Label generated clips in the cut script ("generated": true) and in the description where the platform asks; keep the prompt and model in a sidecar next to the clip. In Claude Code, paste the prompt as a message.
Steps
- Walk the five questions for every MISSING row of your cutaway list; most of a personal film's gaps resolve at question 1 (film it next time, or cut the line).
- For the rows that reach question 4, write the generation prompt from the sentence itself: literal, faceless, in the style lock, and generate 2 takes.
- Put generated clips through the same contact-sheet check as real ones: does it show what the line says?
Tips
- Explainers and product videos reach question 4 constantly; a vlog almost never does. Know which film you are cutting.
- Faceless is what makes generated b-roll safe: no character consistency problem, no likeness question, and it cuts next to real footage without a visible seam when the style lock matches the grade.
| works with | Claude Code: Runs as a skill or a prompt in the terminal; can read your files and write the result. · Veo: Google text-to-video / image-to-video prompts. · Kling: Kling text/image-to-video prompts. · Runway: Runway Gen prompts and camera controls. · By hand: No tool needed: a checklist or a technique. |
| cost | Free. The video or image model you paste it into has its own pricing. |
| 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 | generated-vs-real-broll-decision-rule.md |
Related
Generate a 10-clip b-roll pack for a talking-head video
One agent prompt that reads your transcript and writes ten generation prompts for cutaways that show what is being said, each tagged with the sentence it covers.
Style lock: one paragraph that makes every clip look like the same film
A reusable style paragraph (stock, palette, grain, light rules) appended to every generation prompt so a series cuts together instead of looking like ten different ads.
Before you post a generated video: disclosure and rights checklist
Six checks for AI-generated clips before they go on TikTok, Reels, Shorts or a client's site: labels, likeness, music, logos, model terms and provenance.
Cutaway matching: sentence to a literal visual, same place and time
A prompt that reads each sentence of a transcript and names the one picture that shows it, with the place and time it must come from, so the shot list is a list of things, not moods.
From a sentence to five search queries and one generation prompt
A prompt that turns each line needing a cutaway into five queries for your own footage or a stock site, ordered from literal to loose, plus one generation prompt as the fallback when nothing is found.
B-roll shot list from a transcript: the table to film or generate from
An agent prompt that reads a transcript and returns a shot table (timestamp, line, shot, source: have / film / generate) for a camera or a video model, grouped so each location is filmed once.