Wunmi’sAI STARTING POINT
Menu
Create · Images

AI image generation

Turn an idea into a clear visual brief, use references carefully, and improve an image by checking what it actually shows.

15 minutes + a creative exercise · Beginner first · Optional Go Further

You describe; the tool interprets

Image generation makes an image from a written description, reference material or both. It can help explore layouts, campaign concepts, backgrounds and visual styles. It may also change a face, product shape or written label without your permission.

Your job is to give direction and evaluate the result. A longer prompt is not automatically a better prompt. Start with the subject, purpose and a few visual decisions.

Pick an image-capable tool from the resource guide. These examples are tool-neutral; the controls, reference support, limits and editing options depend on the product. No generated outputs are simulated on this page.

From a basic request to art direction

Basic

“Make a nice photo of a tote bag.”

The tool has to choose the setting, light, framing and purpose.

Directed

“A plain cream canvas tote upright on a light wooden desk, shown from a three-quarter angle. Soft window light from the left, a quiet warm mood, a notebook beside it for scale. Vertical composition with clear space above for a heading added later.”

Subject is what we see. Setting is where it sits. Composition is how elements are arranged. Framing decides how much is visible; perspective is the viewing angle. Lighting affects shadows and mood.

“Close-up” is useful for stitching; “wide shot” gives surroundings. “Eye level” feels different from “overhead”. Choose these because they serve the message, not to make the prompt sound technical.

  1. Choose the subject
  2. Place it in a scene
  3. Set framing and light
  4. Check against the brief

Try a product concept, step by step

  1. Prepare: choose a fictional unbranded tote, or a product reference you own and have permission to upload. Decide on a vertical or square output before starting.
  2. Open image creation: select the image mode in your chosen tool. If it supports references, attach the image using its upload control and state what the reference should preserve.
  3. Generate one concept: paste the brief below. Check generation cost or credits before submitting.
  4. Inspect: compare handles, seams, proportions, lighting and empty heading space with your brief or original product.
  5. Revise one thing: use the tool’s edit function if available, or revise the prompt. “Keep the scene and bag; move the notebook farther right” is easier to assess than “make it premium”.
  6. Finish: add accurate text and logos in an editor. Do not rely on generated lettering for packaging or legal labels.

No image tool available? Sketch the composition as rectangles, label the light direction and use the checklist below. You are practising the same planning skill.

Product concept prompt

Create a product concept of a fictional plain cream canvas tote standing upright on a light wooden desk. Three-quarter view, soft window light from the left, quiet warm mood. One closed notebook beside the bag for scale. Vertical composition with clear space above for a heading I will add later. Plain unbranded surfaces. The bag is the main subject.

Focused edit request

Keep the bag, camera angle and lighting from this image. Change only the notebook position: move it farther to the right so it does not cover the bag. Preserve the existing proportions and seams.

Check your result

For your own review. These ticks are not saved or graded.

Tell the tool what the reference is for

A reference can guide a product’s appearance, a composition or a style. Those are different requests. State which features to preserve and which may change. Not every tool has separate reference controls.

For example: “Use photo A for the bag’s shape and strap placement. Use photo B only for the soft lighting and uncluttered background.” Label files clearly and check that the tool received both. Avoid references that contain private information.

Similarity is not guaranteed. If the result must be an exact record of a real product or person, use the original photograph and conventional editing. Consent to take a photograph does not automatically mean consent to upload it to an AI service.

Reference instruction

I have permission to use this reference. Preserve [specific subject features]. Use it as a reference for [identity / product / composition / style], not as a request to copy every element. Change only [specific change]. If your tool cannot preserve these details reliably, tell me the limitation.

Four briefs you can adapt

These are concept prompts, not photographs of real results. Replace the bracketed details thoughtfully. For a real person or product, verify permission and accuracy before use.

Professional portrait concept

Create a concept for a professional portrait of a fictional adult educator. Eye-level, waist-up framing, relaxed posture, plain burgundy jacket, softly lit warm-white studio. Calm and approachable, natural skin texture, uncluttered background. This is a fictional concept, not an actual person or credential.

Product photography direction

Use my permitted product reference to create a clean concept image of [product]. Preserve its shape, colour and visible features. Place it on a neutral surface with soft side light, a three-quarter angle and realistic contact shadows. Leave space on the right for factual text added later. I will check product fidelity before use.

UGC-style concept

Create a clearly fictional social-content concept: an adult person demonstrating a plain reusable bottle at a kitchen table. Casual phone-camera framing, natural window light, bottle visible, ordinary uncluttered surroundings. No testimonial, results claim or suggestion this is a real customer. I will label synthetic content when sharing.

Editorial campaign direction

Create an editorial concept for a fictional reading campaign. A stack of books on a warm-white bench, one burgundy fabric bookmark, afternoon side light and generous clear space. Overhead framing, restrained palette, tactile paper detail. No generated slogans or readable brand marks; typography will be added in editing.

Common problems and useful corrections

  • Too many props: remove competing objects and identify one main subject.
  • Wrong product details: compare against the reference; do not sell a different feature by accident. Revert to the real photo if needed.
  • Strange hands or edges: inspect at full size; try a simpler pose or local edit, then recheck the entire image.
  • Unreadable text: leave a clean area and add text yourself.
  • Inconsistent series: repeat the same style notes, reference, palette, light direction and framing. Changing all of them at once makes comparison difficult.

Check usage terms and permissions before publication. Synthetic UGC-style content must not pretend to be a real buyer’s experience. Avoid deceptive before-and-after claims.

Go FurtherOptional challenge. Come back whenever you’re ready.

Build a three-image reading campaign: a hero, a close-up and a social crop. Keep the bench, burgundy accent and afternoon light consistent. Change only framing. Compare palette, material, shadows and subject details side by side. Record mismatches and revise the weakest image. You can do this as a three-panel sketch if you do not have generation access.