Download DiffusionBee 1.5.2 (Mac) – Download Free

notes: FP16 The version runs faster, but has a slightly lower build quality. FP32 version yields the best quality.

Recently, there has been a lot of interest in AI-generated artwork based entirely on text cues. Several text-to-image models have been developed for this purpose, each with different capabilities and typically accessible via the web or a command-line interface.

Diffusion bee As a GUI for Stable Diffusion, this is a relatively new project that allows creating images, modifying existing images, repairing and repairing with detailed text prompts. It runs completely offline and takes full advantage of Apple’s M1 and M2 chips.

Generate pictures based on text prompts

You can use any number of keywords to try and get the right image, and there are endless possible combinations. The results can be impressive or weird, but you’ll end up with unique images every time.

In addition to simply describing the content of the image, you can tell the model to use a certain style, drawing tool, color, CGI rendering mode, or even try to emulate the style of a specific artist.

You can generate multiple unique images using the same prompt, customize their dimensions, and change other parameters. After creating an image, you can enlarge it.

Modify an existing image

From the Image to Image tab, you can load an existing image, such as one generated using the same application, and then modify it in various ways. The same text hints are available and you can specify how much the original image should be changed.

The input intensity parameter determines how close the result is to the input image. A high input strength will get you closer to the original picture.

Patching and patching

Healing allows you to make modifications to specific areas of an image using masks. Just like when the image was first generated, text hints are used for new content.

Outpainting, on the other hand, allows you to generate extra content beyond the original dimensions of the image.

Even if you don’t save the artwork, you can still access it in the “History” tab, at least until it’s deleted from there.

DiffusionBee is still under development, but the main features of the AI ​​model are fully functional. It has the advantage of running offline, is optimized for Apple’s in-house silicon, and gives you an intuitive UI to use.