Timed Shot Script for AI Video
You get a clean, timed script where every beat has one shot, one action and one sound. Splitting the clip this way stops the generator blending several actions into a muddle.
Partly workedThe reply gets the structure right: one paragraph, four contiguous segments totalling 12 seconds, consistent woman wording, the MiniDV style notes, and avoid terms stated once. But segments 1 and 4 each combine two actions (sits up and looks; sips and smiles), which breaks the "one main action" and "no competing actions" rules. The "Sound:" labels also depart slightly from the requested "shot type, action, sound" form.
The prompt
Act as a video director who writes precise prompts for AI video generators.
Task: Turn my idea into a second-by-second shot script that I can paste into a video generator.
Steps:
1. Read the idea and decide how to split {clip_length} into 3 to 5 contiguous timed segments with no gaps or overlaps.
2. Give each segment exactly one shot type, one main action and one sound line.
3. Describe {subject} in the same fixed wording every time it appears, so the look stays consistent.
4. Apply the style notes in {visual_style}, and say once what to avoid.
5. Check that the segments add up to the full length and that no segment contains two competing actions.
Output: A single paragraph naming the style and subject, followed by a list in the form "0-3s: shot type, action, sound". Add no commentary.
Idea: {clip_idea}
Change the parts in curly braces. Use it when a video generator keeps merging actions or changing your character mid-clip.
What happened when we ran it
We sent
Act as a video director who writes precise prompts for AI video generators. Task: Turn my idea into a second-by-second shot script that I can paste into a video generator. Steps: 1. Read the idea and decide how to split 12 seconds into 3 to 5 contiguous timed segments with no gaps or overlaps. 2. Give each segment exactly one shot type, one main action and one sound line. 3. Describe a woman in her thirties with short black hair, a mustard raincoat and a small silver nose stud in the same fixed wording every time it appears, so the look stays consistent. 4. Apply the style notes in handheld, early 2000s MiniDV look, soft grain, slightly blown-out window light, no cinematic colour grading, no lens flares, and say once what to avoid. 5. Check that the segments add up to the full length and that no segment contains two competing actions. Output: A single paragraph naming the style and subject, followed by a list in the form "0-3s: shot type, action, sound". Add no commentary. Idea: A woman wakes up early in a small flat, makes pour-over coffee by the window, then takes her first sip and smiles at the camera. Quiet morning mood, rain outside, a kettle and a street tram faintly heard.
The model replied
Handheld early 2000s MiniDV home-video look with soft grain, slightly blown-out window light and no cinematic colour grading, showing a quiet rainy morning in a small flat where a woman in her thirties with short black hair, a mustard raincoat and a small silver nose stud wakes, makes pour-over coffee by the window and smiles at camera over her first sip. Avoid lens flares, stylised colour grading, smooth gimbal movement and any on-screen text or logos.
- 0-3s: Wide handheld shot from the foot of the bed, a woman in her thirties with short black hair, a mustard raincoat and a small silver nose stud sits up in bed and looks toward the rain-streaked window. Sound: steady rain on glass, a faint tram passing far below.
- 3-6s: Medium handheld shot at the window, a woman in her thirties with short black hair, a mustard raincoat and a small silver nose stud pours hot water from a kettle over a pour-over dripper. Sound: kettle water trickling, a soft click of the kettle base, rain continuing.
- 6-9s: Close-up on the dripper and mug in blown-out window light, coffee drips slowly into the mug. Sound: slow drips into ceramic, muffled rain, a distant tram bell.
- 9-12s: Medium close-up, a woman in her thirties with short black hair, a mustard raincoat and a small silver nose stud takes her first sip and lowers the mug into a smile at the camera. Sound: a quiet sip and a soft exhale, rain fading under a faint tram rumble.
Run on Claude Sonnet 5.5, 30 Oct 2026. Automatic check: 3 out of 5.
Mentioned for
ChatGPT, Claude, Gemini, Seedance
Where it came from
Found on GitHub, shared by LearnPrompt: LearnPrompt/awesome-seedance.
This is our own version of the idea, written and tested by us. The original may differ.