Describe the picture you want to see

Create new visuals, edit uploaded images and iterate conversationally—without translating your idea into image-editing software steps.

GenerateEditIterate

A strong image request

👤SubjectWhat must appear?

🧭CompositionWhere should it sit?

🖌️StyleHow should it feel?

📐FormatRatio, background, text

Prompt the outcome, not the software

Name the subjectStart with the person, object or scene.

Set compositionDescribe placement, framing and empty space.

Add visual directionSpecify style, color and important details.

Define the formatGive aspect ratio, background and exact text.

Create a 16:9 flat illustration of a manager reviewing an AI report. Use a warm off-white background, add no text, and leave generous empty space on the right for a slide title.

Edit or regenerate?

🩹

Edit

Use when most of the image is right and one object, detail, label or background needs to change.

🔄

Regenerate

Use when the composition is fundamentally wrong or you want a materially different direction.

🔍

Review

Check exact text, logos, product details and anything factual. A convincing visual can still contain subtle mistakes.

Where image creation still needs human judgment

Image generation is probabilistic. A clearer prompt improves consistency, but it does not remove the need to inspect the final asset.

🔤

Check exact wording

Generated text has improved, but spelling, punctuation and small labels can still be wrong. Review every word when the image will be published or presented.

🏷️

Protect brand and product accuracy

Logos, brand assets, product details and factual diagrams should be compared carefully with authoritative references. A polished-looking image can still invent details.

🖼️

Preserve what should not change

When editing, explicitly state which areas must remain unchanged. The selection tool is a guide rather than a perfect mask, so compare the result with the original before accepting it.