Cute Rainbow Style Mix

Update: I've tried to remake this models using a rebuilt data set, the new versions can be found here, https://civarchive.com/models/2301339/cute-mix-pony-redux

This was mostly because I used broken merges, and didn't train these models properly but have since gotten better at the process.

The new versions don't capture the exact styles, and you do need to prompt more for colorful or rainbow effects, but overall they give better outputs. They also no longer include my BerryBrite or Care Bears Loras in the mix, so they can work better with those.

I also retrained the Cute Mix Illustrious Loras for Pony in the future, and those are posted with this new set.

NOTE: I do plan to remove these models in the near future, as the new set is intended to be a permanent replacement.

This is a blend and merge of some various art styles and some of my previous Loras, meant to capture a cute/chibi character aesthetic.

Cute Styles

Version 1 is a merge of my Care Bears Lora, https://civarchive.com/models/396604/classic-care-bears-pony and

Berry Brite v1 Lora, https://civarchive.com/models/369367?modelVersionId=412745 as such they may give overbaked results when used together.

Version 2, 3 and 4 only include the Berry Brite Lora in the merge, so the Care Bears Lora should work fine with them.

Sample images of these models being used with the Care Bears and BerryBrite loras are below.

Because it includes these Loras in the merge, it's able to use the trigger words from them to give different outputs, however it doesn't capture the styles of those two Loras.

Rainbow Styles

Version 1, This is the first test of the new dataset, so it's not finetuned quite yet. However, it had a cute enough style already so I figured I'd share now.

Version 2 & 3 are two sides of a split in the data set, v2 had the more vibrant, solid, modern 2D animation styles, while v3 had the more digitally painted or hand-drawn styles.

I have noticed that v3 of this set has some issues working certain other loras, however setting the weight to 0.75 or lower seems to work.

Sample Images

First example is a base set, with only the trigger word 'cute'.

Cute Style v1-4

Rainbow Style v1-v3

Rainbow Style v4

(I don't have any plans for future versions)

The Cute Style models were all merged with my BerryBrite v1 Lora (https://civarchive.com/models/369367/berrybrite-pony), and v1 was also merged with my Care Bears Lora (https://civarchive.com/models/396604/classic-care-bears-pony), so using these with those loras could potentially lead to some overbaking in your results. Though from my testing the effects are minor if anything. So, I've included sample images that use both loras together to show my results, using the tags from these two loras.

I've also included sample images for Rainbow Style v1 even though it wasn't merged with these two loras, to show some other styles you can get from them.

Cute Style v1-4 with BerryBrite v1, using the Tkd and TKid tag.

Rainbow Style v1 with BerryBrite v1, using the Tkd and TKid tag.

Cute Style v1-4 with BerryBrite v1, using the Rbrt and RBrite tag.

Rainbow Style v1-4 with BerryBrite v1, using the Rbrt and RBrite tag.

Cute Style v1-4 with BerryBrite v1, using the Sbry and SBerry tag.

Rainbow Style v1-4 with BerryBrite v1, using the Sbry and SBerry tag.

Finally, the Care Bears Lora was only merged with Cute Style v1, and the sample image below shows they have more signs of overbaking when used together. However, the other loras will of course work fine with Care Bears.

Cute Style v1-4 with Care Bears, using the CrBr tag.

Rainbow Style v1-4 with Care Bears, using the CrBr tag.

I know this Lora is a bit all over the place, but I hope some of you can find a use for it.

All of my models may be used for other training and merges.