Inpainting and img2img masking aren't exactly the same thing, though from an end-user perspective, they both work in a very similar fashion -- you paint a part of an image that you want changed, type in a description and voila! With inpainting, you are limited to a very specific model (stable_diffusion_inpainting, based on SD v2). With img2img mask, you can use any model. Also, due to how resource intensive inpainting is, there are not many workers currently running the stable_diffusion_inpainting model, which has resulted in a ton of issues on ArtBot (error messages saying: "No workers available, try again later."). Things should work much better now! Head on over to the inpainting panel and try it out!

This is something I hadn't been doing before now. There are a lot of models now available on the Stable Horde (100 at the time of this writing!). And many of the models are being updated by their maintainers. Some models can drastically change output between differing versions, meaning that the same parameters and seed given to a different version of a particular model will result in a completely different image. In an effort to give more transparency into how your images have been created, ArtBot will now save the version of the model that was created with your image.

When you create a new image request using combinations of various settings (like a prompt matrix, multiple steps, all samplers, etc), image counts can inflate pretty quickly. You now have more visibility into how many images you are requesting from the Stable Horde API, as well as updated kudos costs (total and per image).

Thoughts behind this: Stable Diffusion tends to place more emphasis on words at the beginning of a prompt. If you're using a custom model, you probably want to trigger its special behavior, so placing prompt at front seems to make the most sense. Should this actually be a personal preference / configuration option? Let me know.

Want to see how changing the number of steps affects image output? You can now pass in a comma separated list of values on the create image page and ArtBot will request multiple images. For example, requesting an image with steps of "2,4,16,24" will generate 4 images, each with a step associated with a specific value found within that list.

This is useful in instances where you may have queued up a large number of images on your mobile device and want the process to continue running. This is an experimental feature. if anything looks amiss.

If you have been bitten by the issue where I push an update to ArtBot while you're in the middle of writing a prompt and then lose everything because the page reloads... I'm really sorry! This should no longer happen! Your prompts (and associated image generation settings) will be automatically restored after future server updates.

Absolutely amazing. Thanks to everyone for suggestions, feedback, and using the site. This has been a fun side project and I'm very grateful to Db0 and the rest of the Stable Horde community for creating such an awesome project! Here's to the next million. 😎

It's frustrating to have added all sorts of input for an image and then you click some other link on ArtBot to check something and then all the data is lost when you return. No more! Data is preserved if you move away from the create page. This state gets reset when you create a new image.

Turn your cheesy drawings into awesome AI generated art. This is somewhat in beta as I'm working out some kinks with the painting library (Fabric.js). This should also lay the groundwork for inpainting support once the Stable Horde cluster supports it.

Currently locked to system settings, but I will add an option for user preference in the future.

The thinking here is to simplify the front page as much as possible. If someone has no experience with generative AI art, let them quickly create something.

Relates to future feature that will attempt to detect if job has gone stale and attempt to retry / resubmit

More workers generally means quicker generation times. This is helpful if you want to crank through a number of images.

Add it to your mobile device homescreen for a more app like experience

Due to bandwidth constraints, the Stable Horde sends images as a WEBP file. This isn't always optimal for downloading to your local device. On each image details page, you will see a download button (down arrow).