Turn any image into another, better image.

Create new versions of an image using either commands or descriptions of the final image.

Redraw Image

Advanced

AI model

Instruct what to change

How much of the original image should be turned into noise. At 0.0, the output image will be exactly the same as the input. At 1.0 the output will be completely unrelated to the input. Higher values will result in more change from the input.

Denoising strength

Generate image in response

Output image size (aspect ratio)

Lower values mean more compression and lower quality.

Output JPG compression level

Output format

Output image quality

Higher prompt guidance means your image will look more like the prompt, usually at the expense of lower quality

Prompt guidance

Hide Watermark

Higher image guidance means your generated image will look more like the source image. Try 2.0 to keep a person's face the same. Try lower values like 1.05 if your prompt isn't making enough change.

Image guidance scale

Choose starting image

Write a command like, 'Make the shirt worn by a model.'

Your instruction

Seed

Number of images to generate

Number of inference steps

Generate privately

Describe what you'd like the image to turn into

For a light airbrush, try 3 steps. For more refinement, try 10.

Refiner steps

This is the AI middleman that revises your prompt and actually interacts with the 'gpt-image-1' image generator.

AI orchestrator model

i30 💰:1.0

i66 💰:75.0

i70 💰:125.0

🏆 i67 💰:25.0

i66 💰:12.5

i57 💰:37.5

i49 💰:15.0

i50 💰:1.3

i61 💰:75.0

i55 💰:37.5

i48 💰:11.3

🏆 i44 💰:25.0

i40 💰:5.0

i39 💰:100.0

Sampler name

Scheduler

Strength

Removes imperfections from the generated image.

Use SD XL Refiner

Results

Image-to-image results for: refine it. Generated by diffusers/sdxl-instructpix2pix-768

AI Redraw Image Results: refine it

Results

Request another