Start from a picture, not a blank prompt and say what to change

Upload the image you want the model to work from, describe what should change or carry over, and choose which model answers. The reference does the explaining a prompt cannot.

10 free credits every day - no card required

Made with these exact prompts

A curated set of results. Hover any tile and try the prompt yourself.

Example 1 generated with Runway Gen-4 Image TurboRunway Gen-4 Image Turbo 1c

Change the honey to bright white milk and the wooden dipper to brushed stainless steel, leaving the drop's exact shape, the falling thread, the hard side light and the soft jar rim below completely unchanged.

Example 2 generated with FLUX Kontext ProFLUX Kontext Pro 1c

Change the scene from amber overcast day to clear night: darken the sky to deep blue-black, light every window in the grid in warm gold, add red aircraft beacons on the tallest towers, and leave the street layout and camera exactly as they are.

Example 3 generated with Runway Gen-4 ImageRunway Gen-4 Image 2c

Replace the dark rehearsal hall behind the drummer with a vast open air temple courtyard at golden hour, adding distant stone pillars and a crowd seated on the steps, keeping the drummer, the drum and the hard overhead light exactly unchanged.

Example 4 generated with Seedream v4.0 EditSeedream v4.0 Edit 1c

Convert the image to a detailed scientific watercolor illustration in the style of a Victorian botanical plate, with fine ink outlines, hand-painted color washes, and cream paper texture visible in the background.

Example 5 generated with Seedream 5.0 LiteSeedream 5.0 Lite 1c

Your product, painted the size of a building

Example 6 generated with Nano Banana LiteNano Banana Lite 1c

One packshot, your headline, a finished ad

Why a reference beats a longer prompt

Upload a picture and write a prompt: the model reads both, not just the words
Hold a subject, a pose, a palette or a style steady across a whole set
Several reference-reading models to choose between, so a stubborn look has a second opinion
Add more references when one is not enough to pin the look down
Run it again on the result until the version you wanted arrives
Your upload is processed to produce the result and never used to train a model

Use cases

  • Keeping one character recognisable across a series of images
  • Re-shooting a product in a setting you never had access to
  • Matching a brand palette a written prompt keeps missing
  • Turning a rough sketch or a screenshot into a finished picture
  • Restyling a photo you already own instead of starting over
  • Trying the same reference through several models before choosing one

Image to image, answered

What is image to image?

You give the model a picture as well as a prompt. The picture carries what a sentence struggles with, such as a face, a product, a palette or a composition, and the prompt says what to do with it.

How many reference images can I use?

It depends on the model. Each one declares how many references it reads and the limit is shown beside it in the picker, so the number you see is the model's own rather than a figure written into this page.

How much does it cost?

It runs on credits, and the credit cost is shown before you run it. New accounts start with free credits and there is a free allowance every day.

Can I use the results commercially?

Yes, images generated on Arteza come with a commercial license for business use.

Start from your own picture, free

10 free credits every day - no card required