Timed Shot-by-Shot Video Prompt Writer
It gives you a ready-to-paste video prompt that sets the look first, then breaks the action into timed shots. Video models cope better with clear timings, camera moves and one action per shot than with a loose paragraph.
Worked in our testIt followed the format: a style line with the total length, four timed segments, shot type, movement, setting, action, a lighting or motion detail, spoken lines or cues, and a closing image. Minor flaws: segments 2 and 3 pack in several moves (a push-in, then a whip-pan; a rack focus) and the keeper does a lot in 4 seconds, which makes some shots hard to film in the time. The lantern on his belt conflicts with the "single practical lamp". The style line is also long, and the prompt wanted a single line.
The prompt
Act as a film director and storyboard writer who prepares prompts for AI video generators.
Task: Turn the idea below into one shot-by-shot video prompt in a chosen visual style.
Idea: {scene_idea}
Visual style: {style_reference}
Total length in seconds: {video_length}
Steps:
1. Open with one line naming the visual style, lighting, colour palette and mood, using concrete film terms.
2. Split the total length into three or four timed segments, each with a time range such as [00-05s].
3. For each segment, name the shot type and camera movement, then describe the setting, the subject's action and one lighting or motion detail.
4. Add a short spoken line or emotional cue in any segment where a character reacts, and make the final segment the climax or closing image.
5. Check that every segment can be filmed in its time and that the shots flow from one to the next.
Output: Only the finished prompt, starting with the style line and total length, followed by the timed segments. No commentary.
Change the parts in curly braces. Use it when a vague idea keeps producing messy clips from a text-to-video tool and you want tighter control over pacing.
What happened when we ran it
We sent
Act as a film director and storyboard writer who prepares prompts for AI video generators. Task: Turn the idea below into one shot-by-shot video prompt in a chosen visual style. Idea: A lone lighthouse keeper climbs the spiral stairs as a violent storm hits the coast at night. He reaches the lamp room just as the light fails, relights it by hand, and a stranded fishing boat below turns towards safety. Visual style: 1970s British drama, grainy 35mm film, muted greens and greys, low-key lighting, handheld camera, tense and quiet mood Total length in seconds: 15 seconds Steps: 1. Open with one line naming the visual style, lighting, colour palette and mood, using concrete film terms. 2. Split the total length into three or four timed segments, each with a time range such as [00-05s]. 3. For each segment, name the shot type and camera movement, then describe the setting, the subject's action and one lighting or motion detail. 4. Add a short spoken line or emotional cue in any segment where a character reacts, and make the final segment the climax or closing image. 5. Check that every segment can be filmed in its time and that the shots flow from one to the next. Output: Only the finished prompt, starting with the style line and total length, followed by the timed segments. No commentary.
The model replied
**Style:** 1970s British television drama, grainy 35mm film stock with visible halation and heavy grain, muted greens and greys, low-key lighting with deep shadows and a single warm practical lamp, handheld camera, tense and quiet mood. **Total length: 15 seconds.**
**[00-04s] Medium-tracking shot, handheld, following from below and behind.**
Inside a cramped stone lighthouse tower at night, a weathered keeper in an oilskin coat and wool jumper climbs the iron spiral stairs, one hand gripping the rail. Rain-streaked, grey-green walls close in around him. Thunder rumbles, and a swinging oil lantern on his belt throws a shifting, jittery glow across the damp stone. He breathes heavily. *Spoken, low and to himself:* "Come on, old girl. Hold."
**[04-08s] Close-up, handheld, slight push-in on his face, then a quick whip-pan upward.**
He reaches the top and shoulders open the hatch to the lamp room. Through the salt-crusted glass, the storm rages in black and slate-grey. The great lens goes dark mid-rotation, the beam dying and plunging the room into near-total blackness. Only a faint flash of lightning reveals his wide, alarmed eyes. *Emotional cue:* a sharp, held breath of dread.
**[08-12s] Medium shot, handheld, shaky and close, with a rack focus from his hands to the lamp.**
Working by feel, he strikes a match, cupping the flame against the draught. His weathered hands tremble as he lights the wick. The flame catches, and warm amber light blooms across the brass fittings and the glass prisms, cutting through the cold green-grey. Grain flickers in the glow. *Spoken, a rough whisper:* "There."
**[12-15s] Wide shot, slow handheld drift outward through the rain-lashed window, ending on a held frame.**
The restored beam sweeps out across the heaving, steel-grey sea. Far below, a small fishing boat tossed in the swell sees the light, and its bow slowly turns toward the shore. In the foreground, the keeper's silhouette stands motionless against the glowing lens, his shoulders finally easing. The beam passes across the waves in a steady, quiet arc as the frame holds, thick with grain.
Run on Claude Sonnet 5.5, 30 Oct 2026. Automatic check: 4 out of 5.
Mentioned for
Seedance 2.0, Claude, ChatGPT
Where it came from
Found on GitHub, shared by ZeroLu: ZeroLu/awesome-seedance.
This is our own version of the idea, written and tested by us. The original may differ.