Within the image generator, there are five ways to edit a generated image:

These editing features are not supported by all models but can be applied to an image generated on any model. To demonstrate these features, we'll use this example image which we generated on Pencil pro using the Stable Diffusion model:

(Prompt: Generate a fisheye photo of a green and red parrot looking at the camera. Verdant greens and rich colours in background. Bright natural lighting. Lens 8mm.)

Magic Fill (available with Stable Diffusion and Imagen) allows you to brush over an area that you want to edit or replace. For example, in the image below, we want to change the parrot's eye color. To do this, we click and drag the over the eye, and enter a prompt to indicate what we want to change, such as “Red parrot eye, photo, brightly lit”. You'll get two options to choose from and you can always tweak your prompt and regenerate!

You can see the results magnified below:

Smart Erase (available with Stable Diffusion) allows you to erase small details in an image. For example, to remove cracks in the branch from the image below, we have clicked and dragged the cursor over the cracks. If it doesn’t work the first time, don’t worry, stick with it! You may just need to go over the specific area again.

You can see the results magnified below:

Expand (available with Stable Diffusion and Imagen) allows you to either change the aspect ratio of an image, or add extra imagery around the edge by shrinking the image down. Here we've expanded the image from 16:9 to a 1:1 ratio:

Enhance (available with Stable Diffusion and Imagen) allows you to either increase the overall size of the image (e.g. 2x, 3x, 4x); or, if you keep the image at 1x, you can increase detail within the image.

The creativity slider allows you to tell the AI how much leeway you want the AI to have when it regenerates and enhances your image. Increasing the amount of creativity will allow the AI to create a more pronounced change in your image. The detail will be particularly pronounced when you zoom into your image.

Styles (available with Stable Diffusion) allows you to alter the visual style and color palette of a generated image to match an uploaded style reference. You can control the style strength using the slider on the left. In this example, we used a reference image of a parrot to alter the coloring of our image at 100% strength: