AI image & video models,
and how to talk to each one.

The order that works almost everywhere

1. Deliverable — product photo, poster, infographic, logo, colouring page, book cover
2. Subject — who or what, with identifying details
3. Action or pose
4. Setting — where and when
5. Composition — framing, angle, lens, aspect
6. Lighting — source, quality, direction
7. Colour — palette words or hex codes
8. Style or medium — photo, watercolour, vector, 3D, era
9. Text — exact copy in quotes, position, type style
10. Constraints — what must stay, what to avoid

Copy-paste template

A [deliverable] of [subject + key details], [action], set in [setting].
[Shot size], [angle], [lens / depth of field]. [Lighting]. Palette: [3–5 colours or hex].
Style: [medium / art style / era / texture].
Text: "[EXACT COPY]" in [type style], [position], rendered once, no other text.
Keep: [must-haves].

• Be concrete, not hyped. Name materials, light sources and lenses; drop "epic, stunning, 8k".
• Describe what you want: "empty street", not "no cars". Use a negative field only where the model has one.
• Number reference images and give each a role: Image 1 subject, Image 2 outfit, Image 3 style.
• Edits: "Change only X. Keep face, pose, lighting, camera." One change per turn.
• Draft cheap, finish expensive: low quality while exploring, lock the seed or style, then render high.

Shot-list order

1. Shot and camera — size, angle, movement and its speed
2. Subject — the identical description every time
3. Action — one main action, as countable beats
4. Setting
5. Lighting and colour
6. Style or film look
7. Audio — dialogue, sound effects, ambience

• Image first: lock the keyframe as a still, then describe only motion and camera.
• Positive phrasing: "Locked camera. The camera remains still."
• About one or two short lines of dialogue per 4–8 seconds.
• Close-ups hold a face better than wide shots.
• Resolution, duration and aspect go in settings, not the prompt.

Copy-paste template

[Shot size], [angle], [camera move + speed].
[Subject — identical description every time] [action as beats: takes three steps, pauses, turns to camera].
Setting: [where / when]. Lighting: [key source + quality], palette [3–5 words]. Look: [film stock / grade].
Dialogue: [Name] ([tone]): "[one short line]"
SFX: [sound tied to the action]. Ambient: [background bed].

Multi-shot

Shot 1 [0–4s]: Wide establishing, slow drone descent over …
Shot 2 [4–8s]: Medium, handheld follow — [Name: short fixed description] …
Shot 3 [8–12s]: Close-up, locked camera. [Name] (warm): "…"