When to use it
You have a video topic and you need to know BEFORE the shoot exactly what to film and say — shot by shot, so you're not improvising on location. This is an input plan (what to film), unlike an edit plan, which works with footage you already have. The AI's role is director-writer. Output: a shot list with the script, ready to shoot and cut against.
The prompt (copy and paste)
You are a director and writer of short videos. Make me a shot-by-shot script/storyboard.
TOPIC AND IDEA: "<PASTE: what the video is about and what the viewer should understand/do>"
FORMAT: <YouTube horizontal / Shorts 9:16 vertical / Reels> RUNTIME: <e.g. 45 sec>
TONE AND STYLE: <e.g. energetic, straight to camera>
RESOURCES: <what I have: phone camera / screen / graphics; shooting solo, etc.>
Give me the storyboard as a table, shot by shot:
| # | dur. (sec) | voiceover / line | what's on screen (visual) | shot type (close-up/medium/screen/B-roll) | on-screen text |
Rules:
1. Shot 1 is a 2-3 second hook: it lands from the first second (a question, a promise, a conflict), no long intro.
2. One shot, one thought; keep it moving, change shot types, don't let a talking head sit longer than it needs to.
3. Write the voiceover the way people SPEAK, not the way they write — short phrases built for reading aloud.
4. End with a clear CTA shot (what the viewer should do).
5. The shot durations must add up to the given runtime (make the numbers work).
At the end give me: a list of what to film/prepare (shot list), and a note on music and pacing.
Where you filled in details, mark them [assumption]. The lines must be original, not stock phrases.
The technique: a shot-by-shot table of "line → visual → shot type → on-screen text" plus the rule "durations sum to the runtime" produces a shooting document, not a general restatement of the idea.
Filled-in example
Topic: "3 phone settings that instantly improve your videos". Format: Shorts 9:16, 40 sec. Tone: energetic, to camera. Resources: phone only, shooting solo.
Expected AI response: shot 1 (3 sec, close-up to camera) — "Do your videos look cheap? It comes down to three settings"; shots 2-4 — one setting each: line + a screen capture of the phone menu + on-screen text "1 / 2 / 3"; shot 5 (CTA) — "Save this so you don't forget"; total = 40 sec; shot list "record 3 screencasts of the settings + 2 talking heads", music — an upbeat beat, cut on the beat.
Variations
- Ad spot. Add the "problem → agitation → solution → offer" formula to the rules and have it laid out shot by shot.
- Voiceover only. "Give me a clean voiceover script with timings for recording, no visual column" — for when the footage already exists.
- For AI video. "For each shot, give me a text prompt for a video generator (Runway/Veo/Sora class)" — the storyboard as a spec for the model.
Pro tips
- Force the durations to add up to the runtime — otherwise a "40-second" script stretches to two minutes on location.
- Read the voiceover out loud against the timing: if it doesn't fit or sounds "written", ask for a rewrite "the way people talk".
- A storyboard is an input plan (what to film). Once it's shot, hand the footage to an edit plan keyed to timecodes: two different stages, don't mix them up.