What a shot list is
One row per shot: what we see, how the camera sees it, how long it lasts and what we hear. It is the bridge between the story and the prompts. Still prompts come from “what we see”; motion prompts come from “camera”.
The templates below work in Claude, ChatGPT, Gemini or any other chat assistant. Replace the parts in [brackets].
Step 1. The premise
Prompt template
Act as a film director’s story partner. My idea: [one or two lines] Length: [15 / 30 / 60] seconds. Format: [9:16 reel / 16:9 film]. Give me 5 different one-sentence premises in this form: “A [character] wants [visible goal], but [obstacle].” Each one must be something a camera can show without dialogue. Do not write scenes yet.
Pick one and change the words until it sounds like yours.
Step 2. The beats
Prompt template
Premise: [paste your sentence] Break it into three beats: hook, turn, payoff. For each beat give: - what the viewer sees - the question it raises or answers - how many seconds it gets out of [length] The hook must work with the sound off.
Step 3. The shot list
Prompt template
Turn these beats into a shot list of [6–10] shots. Return a table with these columns: shot number | beat | seconds | shot size (wide, medium, close-up, insert) | camera (locked, push-in, pull-out, pan, tilt, orbit, handheld) | what we see, in one concrete sentence | sound (ambience, effect or music cue) Rules: - One idea per shot. - No shot longer than 5 seconds. - Change the shot size from one shot to the next. - Describe the character’s face, hair and clothing once, then keep them identical in every row.
Step 4. Check it like an editor
- Read only the “what we see” column, top to bottom. The story should still make sense.
- The first shot is the hook, not a slow establishing wide.
- No two neighbours share the same size and angle.
- The seconds add up to your length.
- Each shot holds one clear action. “She runs, trips and gets up” is three shots.
Prompt template · the editor pass
Here is my shot list: [paste the table] Act as a strict film editor. Tell me: 1. Which shots can be cut without losing the story. 2. Where two shots in a row look too similar. 3. Any shot that asks for more than one action. Then return the corrected table only.
Then write the prompts
Each row becomes two prompts: a still prompt (what we see, plus the look) and a motion prompt (camera, plus the action). See Lock a look in Midjourney and Camera moves for AI video.
For finished examples, open any film breakdown.