Prompt like a director

Fill in the brackets. Aim for 20 to 40 words.

[Medium and style].
[Main subject, with its age, clothes and materials right next to it].
[Second subject and exact position: left or right, on which surface].
[Setting: place, time of day, weather, background].
[Light: key light direction, colour, shadows].
[Framing: shot size, camera angle, lens, aspect ratio].

A longer version from my research notes. It runs over 40 words, so try trimming it.

Editorial 35mm cinematic film still. A weathered 45-year-old detective in a damp olive-drab trench coat, walking along a wet asphalt street slightly left of frame, adjacent to a retro-futuristic black sedan on the right curb. Dense dystopian city alleyway at night with dark brick facades. Backlit by a pale blue overhead streetlamp with rim light on the coat, contrasted with warm magenta neon reflections in street puddles. Wide-angle full-body shot, low-angle perspective, 28mm lens at f/2.8, atmospheric mist, 16:9 aspect ratio.

Paste into Claude, ChatGPT or Gemini.

You're a film director and prompt writer. I want an image of [your idea], made in [model].
Ask me up to 5 questions about the subject, setting, light and framing. Wait for my answers.
Then write one prompt, 40 words or fewer, in this order: medium, subject, position, setting, light, framing. Keep each object's details next to its noun. Describe only what should appear.

Upload your image and paste your brief in the same chat.

Compare this image with my brief. Tell me the one slot to change, and rewrite only that slot.

Change one thing per turn. Mask the area too if you can.

Change only: [one named thing].
Keep: identity (face, features, skin tone, body shape, pose), geometry, layout, lighting, camera angle, surrounding objects.
Constraints: no added accessories, text, logos or watermarks. Do not change anything else.

Attach your images first, and say what each one is for.

Using the attached [image 1] as the [character, structure or style] and the attached [image 2] as the [location, texture or colour], create [new scene].

Upload your locked frame first, then paste this.

Use my image as the first frame.
Camera: [one move, such as a slow push in].
Subject: [a verb you can see, with a direction].
Scene: [what changes over time].
Continuity: [what stays the same: faces, clothes, light].
Format: one shot, [8] seconds, 16:9.

For Seedance 2.5 and similar models. Give every shot its own start and end time.

[Duration], 16:9. [One sentence on what happens.]
REFERENCES: @name: [what to take from it, and what to ignore].
0 to 4s: [camera position, lens, move]. [Who is where]. BACKGROUND: [what's behind them]. [Action].
4 to 8s: [camera position, lens, move]. [Who is where]. BACKGROUND: [what's behind them]. [Action].
SOUND: [ambience, and any lines, each spoken once].
KEEP CONSISTENT: [faces, clothes, seats, what must not change].

Make it yours: change the details.

35mm film still. An open carton of brown eggs on a supermarket chiller shelf, one egg cracked, beside plastic-wrapped chickens. Cold overhead light. Close-up, eye-level, 50mm, 16:9.

Use after you've locked the first frame.

Slow push in on the carton. The cracked egg rocks and a yellow chick climbs out. Wrapped chickens burst their plastic and flap across the aisle. Same shelf, same cold light.

Make it yours: change the details.

35mm documentary still. A hungry traveller in a blue linen shirt, seen from behind, walks towards a shawarma stall. Stone arches above. Warm afternoon sun. Eye-level, 35mm, 16:9.

Use after you've locked the first frame.

Handheld tracking shot behind the traveller. They weave between shoppers and stop at the stall. The vendor slices shawarma and hands over a wrap. Same blue shirt, same arches.

Paste this into ChatGPT or Claude, with the guide for your model.

Here's the prompting guide for [model]: [paste the link or the text].
Write me a prompt for [your idea].
Ask me questions first if you need to.

Make this first, then attach it whenever that character appears.

Create a 16:9 image: Character reference sheet on a plain grey background. [Who they are: age, hair, clothes, one standout detail]. Full body from the front, side and back, plus close-ups of their face and [the standout detail]. Same person and same outfit in every view.

Nine views of one place, so every shot feels like the same place.

Create a 16:9 image: Location reference sheet, a 3 by 3 grid of views of the same place. [The place: what it is, materials, colours]. Wide, medium and close views from different angles. Same place and same light in every view.

Attach your sheet, then paste this.

Using the attached sheet as the character, create: [Medium]. This exact person, in the same clothes, [what they do], [where]. [Light]. [Framing].