This release brings with it some bug fixes, new refreshes to the UX and several new features, let’s dive in:

New Features

First off, two new popular image to image generators

Pix2Pix

Pix2pix was recently released as a new way to modify an existing images with text. Start with an image, set a prompt and hit generate.

Depth2img

Depth2img works similarly as pix2pix and img2img - start with an existing image, add some text and the model will generate a depth mask and modify your original image using it.

In the future you will be able to use your own depth mask.

United canvas

AI Runner introduces a united canvas. This means that you can switch between tabs and work with the same image. Only txt2img will replace your existing art with a new image - all other tabs will manipulate your existing artwork. Simply highlight the area you want to modify with the active area square and press generate.

Limited CPU support

We don’t recommend it, but you can now generate images using your CPU. Expect very slow times in comparison to GPU. This will be useful for users who want to test the technology but who have not yet been able to acquire a CUDA capable video card.

Stats

Get active memory stats at the bottom of the screen to monitor your GPU and CPU memory usage.

Updates to existing features

Image gallery updates

• The image gallery now loads faster and is more responsive
• Thumbnails in the gallery will show a red border when active

Misc updates

• Canvas dragging In previous versions you could hold right click on your mouse and drag, now you can do this will the middle mouse wheel as well
• Better handling of active area box images will generate wherever you place the active area box
• Better inpainting no more image shifting or replacement of the active area box
• Faster and more reliable you should see less memory crashes, faster model switching, better handling of ckpt and safetensors.
• UI updates Brush and active area box sizes have been moved to a row underneath the toolbar