How to Use Reference Images to Get Exactly the AI Art You Want

Most people use AI image generators by typing a prompt and getting an image. That works. But there is a feature most people skip that transforms the results: reference images.

A reference image tells the AI "make something like this." Not copy it — learn from it. The AI studies your reference for composition, color palette, lighting style, and mood, then applies those qualities to your prompt. Here is how to use references well.

What Reference Images Actually Do

SDXL supports img2img mode. You upload a reference image alongside your text prompt. The AI uses the reference as a starting point rather than pure noise. This means the output shares structural and stylistic qualities with your reference while following your text prompt for the subject.

Style transfer: Upload a photo with lighting you love. The AI preserves the lighting style and applies it to your new subject. Composition guidance: Upload a sketch or wireframe layout. The AI matches the rough composition. Color palette matching: Upload an image with colors you want. The AI adopts the palette.

How to Use References Effectively

Upload your reference image before generating. Tips: use references that show the STYLE you want, not the SUBJECT. If you want a photorealistic portrait with dramatic lighting, upload a photo with dramatic lighting — not necessarily a portrait. Upload at a reasonable resolution. The reference costs 1 extra credit.

The AI image generator gives you 1-4 output images with controls for quality, aspect ratio, and format. For the best results, pair a detailed prompt with a well-chosen reference image.