10 image generation prompts

Composition, lighting, material and subject direction — prompts written for image models rather than chat models. Written for Midjourney, DALL-E, Stable Diffusion, Ideogram. Square brackets mark what you fill in; everything else can be pasted as it stands.

[Subject], [what it is doing or how it is arranged], [setting].

Lighting: [single hard key from the left / soft overcast / backlit with rim].
Camera: [35mm wide, low angle / 85mm portrait, eye level / macro].
Palette: [two or three colours only].
Mood: [one word].

Negative: no text, no watermark, no logo, no extra limbs, no border.

I want [mood] lighting for [subject]. Translate that into physical terms.

Give me the key light's direction, size and hardness; whether there is fill and how much; whether there is a rim or kicker; the colour temperature of each; and what the shadows should look like.

Then write it as a prompt fragment I can paste in.

Studio product photograph of [product] on [surface], seamless [colour] backdrop.

Lighting: large softbox above and slightly front, white bounce card on the right, controlled specular highlight along the top edge.
Camera: 100mm macro, f/8, straight-on at product height.
Composition: product occupies the left third, generous empty space to the right for copy.

Photorealistic, commercial catalogue quality. No props, no text, no hands.

Flat graphic poster in the Swiss international style, containing no typography at all.

Shapes: large, precise, hard-edged — exact circles, arcs, rectangles, repeated bars. One shape dominates.
Palette: exactly three flat colours — [colour one], [colour two], [paper colour]. Flat fills only.
Composition: asymmetric on an invisible grid, generous negative space.
Surface: subtle offset-print grain, faint halftone in one area.

No 3D, no gradients, no shadows, no letters, no numbers. 16:9.

[Subject and setting].

Critical: the [left half / lower third] of the frame must stay visually calm and free of fine detail — a headline will be placed there. No bright areas in that region.

Lighting: [direction], overall mid-to-dark value.
Camera: [lens and angle], wide 16:9 framing.

No text, no logos, no people's faces.

Icon representing [concept], designed to be legible at 24 pixels.

Style: solid single-weight shapes, no gradients, no inner detail smaller than one tenth of the icon, centred with even optical margins. Monochrome [colour] on transparent.

Produce five distinct metaphors for the concept, not five styles of the same metaphor. For each, name the metaphor in one word.

This prompt keeps producing [what goes wrong]:

[paste the prompt]

Tell me which phrase is responsible. Image models weight early words more heavily and ignore negations inside descriptive phrases — say whether either is the problem here.

Then give me a corrected version, changing as little as possible.

First, write a character description I can paste into every future prompt. It must fix: age range, build, hair, clothing with specific colours and materials, and one distinctive detail.

Keep it under forty words and avoid anything the model will interpret differently each time, such as "beautiful" or "friendly".

Then use it in this scene: [scene]. Keep the character block identical and change only the scene.

Look at this reference image: [attach or describe it].

Describe its style concretely so I can reuse it: palette with actual colour names, lighting direction and hardness, level of detail, texture, composition rules, and what it conspicuously avoids.

Do not use subjective adjectives. Then write a prompt that would produce [new subject] in that style.

I need [number] images for [use], covering these subjects: [list them].

First, write the shared style block — palette, lighting, camera, treatment — that every image will use unchanged.

Then write one subject line per image. Keep the subject lines the same length and structure, so nothing drifts. Say which subject is most likely to break the consistency and why.