ChatGPT Prompt for Image Generation

Photo Editing Updated August 24, 2026 by GenPrompto
Prompt
Generate an image of [subject, described specifically -- physical details, not just a category], in [style, named precisely -- e.g. "photorealistic," "flat vector illustration," "watercolor painting"], set in [setting, with concrete detail on location and time of day], with [lighting, described directionally, e.g. "soft light from a window on the left"]. Framing: [shot type, e.g. "close-up," "wide shot"]. Aspect ratio: [e.g. "square" or "widescreen"].

What this does

A from-scratch image generation prompt for ChatGPT -- subject, style, setting -- for illustrations, concept art, or photorealistic scenes.

Generating a brand-new image with ChatGPT — rather than editing an existing photo — works best when you name the subject, style, and setting together, since leaving any of the three vague produces a generic, catalog-style result.

What this produces

A newly generated image matching the subject, style, and setting you specify — created from scratch rather than derived from an uploaded photo.

Variations

Two ready-to-copy versions for different needs.

Illustration/Concept Art Version
Generate an illustration of [subject, described with specific visual detail], in [art style, e.g. “digital painting with visible brushwork” or “flat vector illustration with clean geometric shapes”], set in [setting, described concretely], with [mood or lighting, e.g. “warm late-afternoon light” or “cool blue moonlit tones”]. Include [any specific compositional elements you want, e.g. “a strong diagonal line leading the eye toward the subject”].
Photorealistic Scene Version
Generate a photorealistic image of [subject or scene, described with real specificity], [setting, with concrete location and time-of-day detail], with [lighting, described directionally and by quality — e.g. “hard midday sun creating strong shadows” or “diffused overcast light with minimal shadow”], shot as a [framing, e.g. “wide establishing shot” or “tight close-up”]. Include realistic environmental detail (texture, reflections, atmospheric perspective) appropriate to the setting.

Who this is for

For building an image from nothing — no existing photo involved — where subject, style, and setting are the three things actually worth specifying.

Example Output

Naming subject, style, and setting together produces a coherent new image on the first try — leaving any of the three vague tends to produce a generic, catalog-style image that doesn’t match what you pictured.

Tips for better results

  • Name a specific art style (photorealistic, watercolor, 3D render) — ChatGPT’s default style guess is often not what you pictured.
  • Describe the setting and mood, not just the subject — it fills in the rest of the scene coherently.
  • Mention framing (close-up, wide shot) to control composition.
  • If you’re not starting from an existing photo, say so clearly — this avoids ChatGPT assuming you want an edit instead.

What didn’t work as well

Naming only the subject with no style or setting, e.g. just “a dragon.” ChatGPT fills every gap with its own default choices, which rarely matches what you pictured — naming style and setting together is what removes that guesswork.

Related: try ChatGPT Prompt for Realistic Photos or ChatGPT Trait Prompt for Photo Editing for a different angle on this.

FAQ

Can I use this to edit an existing photo instead?

This prompt is built for generating new images from scratch — for editing an existing photo, use the dedicated ChatGPT Prompt for Photo Editing page instead.

How many images does ChatGPT generate per request?

Typically one image per request — ask for a few variations if you want options to compare.

Can I request a specific aspect ratio?

Yes — mention the aspect ratio (square, portrait, widescreen) explicitly in your prompt.

Related