Zabadu's Soft Styles

Similar to the older collection of style Loras I made for Pony, this will be a new collection of art styles for Illustrious.

Unlike the Pony models, which were mostly merges of various style loras (which was done poorly by me, so they have their own issues), these new Loras for Illustrious are being trained on datasets generated from mixing styles into something I like.

This gives me a place to put art style loras I make that are simpler, and don't need improvements later, so I'll update this as I create new style variants.

These Loras will be trained without trigger words, as the purpose is to capture the general art style, rather than a specific aspect of something. So I won't go into much detail for prompting, I'll just give short descriptions below, and you can view the sample images for the prompts used.

Soft Comic Style

This style is based on various western comic art styles.
It combines soft details and colors against sharp black line-art and shadows.

Soft Toon Style

Though I've already made various cartoon inspired styles, with my BerryBrite and Cute Mix loras, this one is based more on the styles of my Zabadu's Mix Anime/Toon checkpoint merges. It features bright colors, with soft shading.

Soft Drawing Style

This one is partly inspired by my Zabadu's Mix Drawn Detail checkpoint merges, but I added a bit of extra color influence to make it more vibrant.
It has thinner line-art, with some more sketchy lines, shading shading, a vibrant color palette and a pop of shiny highlights.

Flat Toon Style

This one is almost a mix of the Soft Comic and Soft Toon styles.

I tried to step back from the more modern toon-ish style of the previous lora, and get some more classic style animation, with better proportions and flatter colors, and a slight hint of cel shading. This just gives a nice variant of the other two styles.

(Future additions will be placed here)

These Loras are trained using Illustrious v2, and the images are generated using Fooocus with Wai Illustrious. For the base samples I won't upscale or inpaint, showing the details as the lora outputs.

As with all of my models, you may use them in your own training, merges, or anything else.