Style Reference

Upload a single image to lock the aesthetic, finish, and mood across generations—without specifying every adjective.

What a style reference does

A style reference pulls finish, color grade, grain character, and atmosphere from the uploaded image. It is the fastest way to match a brand look without a long negative prompt list.

Google’s Nano Banana / Gemini image docs treat reference images as first-class inputs: you can combine many references (up to 14 in advanced workflows) with text to steer composition—see Gemini API — image generation.

When to use it

  • You have a brand palette and a few reference photos.
  • You want consistent output across a campaign of 50–100 images.
  • You have a photographer’s lookbook you want to scale.

How to prepare the reference

  1. Crop to the region that carries the style (texture, grain, or grade).
  2. Strip metadata if you want no subject identity bleeding through.
  3. Use at least 1024 px on the short axis for the style to register.

Combining with text

Specify what stays and what can change:

“Style reference: [uploaded image] — keep the film grain and warm shadows; change the subject to a running shoe on asphalt.”

Style transfer example: city photo transformed to Van Gogh style

Limits

If the reference has strong subject content (specific faces, logos), the model may blend those unless you use negative prompts or describe the subject textually with more detail.

See also