Guide to Inpainting with Stable Diffusion

What is Stable Diffusion inpainting?

In image editing, inpainting is a process of restoring missing parts of pictures. It's most commonly used to reconstruct old deteriorated images and remove cracks, scratches, dust spots, or red eyes from photographs.

However, with the power of AI and the Stable Diffusion model, inpainting can be used to achieve more than that. For example, instead of just restoring missing parts of an image, it can be used to render something entirely new in any part of an existing picture. Your imagination is the only limit.

The inpainting feature is available in AI Canvas and Image Editor. This guide will explain how to use it effectively with the first of these tools.

How to use inpainting in AI Canvas?

Let's look at a simple example of inpainting with AI Canvas. We will use Johannes Vermeer's "Girl with a Pearl Earring" as a base.

Step 1. Create a project and upload an image

First, create a new project and upload the picture onto the canvas. You can upload an image to AI Canvas from:

• you device: by clicking the empty canvas and choosing a file from the list, or by dragging a file to the canvas;

• AI Generator: by clicking the three dots icon in the corner of an image and selecting "Open in AI Canvas";

• the Gallery: by clicking an image, then the "Reuse image" button and selecting "Open in AI Canvas".

Step 2. Erase elements you want to replace

Just click on the Eraser icon at the top, select the brush size, and drag over the part you want to be replaced, just like in any other graphics editor.

But remember to ensure that the object is removed entirely, as even a couple of pixels present can guide Stable Diffusion in the wrong direction.

We will try to use inpainting to replace the girl's head covering with red hair, so let's erase it.

Step 3. Move the Generation Frame and write a prompt

Now, we can use AI to render something new in that empty spot. To do this, move the 'Generation Frame' so that it covers the erased part.

Ensure you select "In/Outpaint" mode and describe what you want to see. Before generation, you can change the batch size and other generation parameters (the model, seed, CFG scale, negative prompt, and steps).

When you're done, click "Generate".

By default, AI Canvas provides you with four images to choose from. You can accept what you like the most and continue editing, or cancel and start over to get more examples.

Finally, we can compare the initial image with our outcome. Stable Diffusion Inpainting not only rendered exactly what we wanted to see in the missing spot but also seamlessly matched the style of a source image.

"Girl with a Pearl Earring"

"Girl with Red Hair"

We hope our guide helped you understand how inpainting works with Stable Diffusion. We're sure you will use what you've learned to create something exceptional! If so, please join our Discord and share it with our amazing community.

Frequently Asked Questions

How does Stable Diffusion inpainting work?

Inpainting uses AI to fill in parts of an image. You can create new content that blends seamlessly with the existing image by erasing unwanted sections and describing what you want in their place.

What types of images can I use for inpainting?

You can use inpainting with a wide variety of images, including photos, digital art, and traditional artwork.