Guide

How to create an AI prompt from a photo

Reverse-engineer any image into a reusable prompt for ChatGPT, Midjourney, Gemini, or Claude — even if you've never written a prompt before.

Great AI images almost always start from a great reference. If you have a photo whose look you want to recreate — a lighting style, a composition, a mood — you don't have to guess at the words. You can hand the photo to a multimodal AI, get a first-pass prompt, then shape it into a template you can reuse forever.

This tutorial walks through the exact five-step process.

1. Pick a clear reference photo

Choose an image where the subject, lighting, and composition are obvious. Cluttered or low-resolution photos produce vague prompts. If you want the AI to imitate a specific style (film grain, isometric illustration, pixel art), pick a photo where that style is unmistakable.

2. Describe what you see, in six buckets

Before touching any AI tool, write one line for each of these:

Subject — who or what is in frame
Setting — where they are, background details
Lighting — soft window light, harsh sun, neon, studio
Camera — close-up, wide, low angle, drone
Style — photorealistic, cinematic, illustration, 3D render
Mood — cozy, dramatic, minimalist, chaotic

This is the skeleton of your prompt. Even if the AI writes the final wording, the structure is yours.

3. Use an image-to-prompt tool

Upload the photo to any multimodal model:

ChatGPT — attach the image and ask: "Describe this image as a prompt I can paste into Midjourney."
Gemini — same flow, particularly good at photographic detail.
Claude — great for capturing mood and stylistic nuance.

Ask for the format you actually need: a Midjourney prompt reads differently from a DALL·E prompt or a Nano Banana prompt. Specify the generator by name.

4. Refine and templatize

Copy the AI's output into a text editor and:

Delete filler words ("beautiful", "stunning") that add no information.
Replace the changeable parts with [placeholders] — [subject], [time of day], [color palette].
Add the technical suffix your generator expects (aspect ratio, style version, quality flags).

You now have a template, not a one-off.

5. Test, compare, iterate

Run the prompt through the generator you targeted. Put the AI output next to the original photo and note what's off — usually lighting or camera angle. Add one specific correction per iteration ("shot on 35mm film", "golden hour side light") and re-run. Three or four rounds is usually enough.

Ready-made prompts

The Prompt Library collects tested image, video, and text prompts you can copy in one click — including several photo-to-prompt starters for portraits, product shots, and cinematic scenes.