AI image prompts for generation and photo editing

Start with a clear task

A useful image prompt defines a result you can inspect. Name the subject, where it appears, the light, the framing and the style. For a fictional product concept, describe a blue ceramic cup on a pale surface rather than asking only for a beautiful picture.

Generation starts with a description. Editing starts with an uploaded image you have permission to change. When editing a product photo, name the change and the details to keep: shape, color, label, crop and lighting. The model may still alter them, so compare the result with the original.

The prompt builder below assembles text locally. It does not upload an image, call an AI model or spend credits. Image generation and editing in Image Studio use image credits from a credit pack or monthly plan. Creating an account does not include a free image attempt; check the balance and displayed cost before submitting.

The structure: subject, setting, light, lens, mood

Work through five slots in order. Subject (who or what, and what they're doing). Setting (where, and what's around them). Light (direction, quality, time of day — this does more work than anything else). Lens and framing (wide, close, eye-level, from above). Mood (the feeling, and any style reference).

Lighting is the highest-leverage slot. "Soft light from a window on the left, late afternoon" changes an image more than any number of quality adjectives.

• Weak prompt

  A beautiful photo of a coffee cup, 8k, stunning, highly detailed.

• Strong prompt

  A ceramic coffee cup with steam rising, on a worn wooden table beside a folded newspaper. Soft morning light from a window on the left, long shadows falling right. Shot at eye level, 50mm, shallow depth of field with the background falling off. Quiet, unhurried mood. Natural color, slight film grain.

• Why it works

  Every clause narrows the space of possible images. The weak version's adjectives narrow nothing — "stunning" and "8k" don't tell the model where the light comes from or how close the camera is, so it picks its default.

Build an image prompt that says what you mean

Edit this fictional example or start with your own subject. This builds text in your browser; it does not create an image or use credits.

Review the result at full size. Exact text, product details and likenesses need your own check.

An account is required. Check your available image allowance and the job cost in Image Studio before generating or editing. One-time packs and monthly plans provide image credits.

• A fictional product concept

  A blue ceramic cup on a pale studio surface. Soft light from the left. Square composition at eye level. No writing or logo.

• An original coloring page

  An original illustration of a cat sitting in a garden beside large flowers and leaves. Black outlines on a white background. Portrait composition with a clear margin. No shading, no small intricate details, no text.

The prompts

• Product on white

  [Product] centered on a seamless white background, studio lighting with a soft key from the upper left and subtle fill, no harsh shadows, slight reflection beneath. Shot at eye level, 85mm, sharp throughout. Clean e-commerce product photography.

• Product in context

  [Product] being used by [person] in [realistic setting]. Natural light from a window on the [left/right], late afternoon. Shot slightly above eye level, 35mm, subject sharp with background gently blurred. Candid, unposed, documentary feel.

• Portrait

  Portrait of [subject description], [expression], looking [direction]. Soft directional light from the left at 45 degrees, gentle shadow on the opposite cheek. Shot at eye level, 85mm, shallow depth of field. Neutral background. Natural skin texture, no retouching.

• Landscape

  [Landscape description] at [time of day]. Light [direction and quality — e.g. low golden sun from behind camera]. Wide shot, 24mm, deep focus from foreground to horizon. Foreground element: [detail] for depth. [Weather/atmosphere].

• Food

  [Dish] on [surface], shot from [45 degrees / directly above]. Soft diffused window light from the back-left, creating gentle highlights on the food's texture. 50mm, shallow depth of field. Props: [items] arranged casually, not styled to perfection. Warm natural color.

• Interior

  [Room type] interior in [style]. Natural daylight from [window position], soft and even. Wide angle, 24mm, shot from corner at chest height, verticals straight. Materials: [textures]. Lived-in, not staged — [specific detail of use].

• Illustration

  Flat vector illustration of [subject]. Limited palette of [3-4 colors]. Bold simple shapes, minimal detail, no gradients, clean geometric composition. Even lighting, no shadows. Modern editorial style.

• Character concept

  Character concept art: [description including build, age, clothing, distinguishing features]. Neutral standing pose, three-quarter view, even studio lighting so all details are readable. Plain background. Consistent design suitable for reference.

• Hero background

  Abstract background for a website hero: [color palette], soft gradient with subtle organic texture, no recognizable objects or text. Even lighting, low contrast so overlaid text stays readable. Wide 16:9 composition with the left third kept visually quiet.

• Editorial photo

  Editorial photograph illustrating [concept]. [Subject and action]. Available light only, [direction and quality]. Shot at 35mm, subject placed off-center following the rule of thirds. Documentary style, natural color, no obvious staging.

• Consistent series

  [Subject description — keep this text identical across generations]. [Vary only this: setting/pose]. Same lighting: [specify]. Same lens: [specify]. Same color treatment. Consistent character design.

• Mockup scene

  [Device] on [surface], screen visible at a slight angle. Soft overhead light with a gentle gradient falloff. Shot from above at 30 degrees, 50mm. Minimal props: [1-2 items]. Muted background so the screen is the focus.

• Texture / pattern

  Seamless repeating pattern of [motif]. [Palette]. Flat even lighting, no perspective, tileable edges. [Style — geometric/organic/hand-drawn]. Suitable for background use at small scale.

• Moody atmosphere

  [Scene] at [time]. Single light source: [specify — streetlamp, screen glow, moonlight] creating strong directional shadows. Low key, deep shadows retained. 35mm, subject small in frame to emphasize environment. [Weather/haze]. Desaturated except [accent color].

• Fixing a generation

  Same as before, but change only: [one specific thing]. Keep the subject, lighting direction, framing, and color treatment identical.

• Edit a product onto a white background

  Edit my authorized product photo. Replace only the background with a plain white studio surface. Preserve the product shape, color, cap, label, proportions and original crop. Keep a subtle natural contact shadow. Do not add lettering, logos or new product details. I will compare the result with the original before publishing.

• Soften a harsh shadow

  Edit my authorized photo. Reduce the harsh shadow on [specific area] while keeping natural light direction and texture. Preserve the person or object, colors, crop and background. Do not invent detail hidden in darkness or change the face. Keep the result natural rather than flattening every shadow.

• Remove one distracting object

  Edit my authorized photo. Remove only [object and precise location, for example the red mug at the back-left of the table]. Reconstruct that small area consistently with the surrounding surface and light. Preserve all other objects, the main subject, text and crop. Do not add anything. I will inspect the repaired area at full size.

Common mistakes

• Stacking quality adjectives

  "8k, ultra detailed, masterpiece, stunning" narrows nothing. The model already aims for quality — those tokens compete with the ones carrying real information.

• Not specifying lighting

  State where the light comes from and whether it should be soft or harsh. This gives you a concrete part of the brief to check; it does not guarantee a particular result.

• Changing everything at once

  When iterating, change one variable per generation. Otherwise you can't tell which change caused the improvement.

• Describing the scene again when editing

  When refining an existing image, re-describing everything invites the model to change things you liked. Name only the delta.

Frequently asked questions