Yet Another Anime Style (but this time for Flux 2 K 9B!) v1:

I created this LoRA for Flux 2 Klein 9B for a project I was working on to animate a bunch of existing real-life photos and was faced either with downloading a better suited anime modern checkpoint model , like Illustrious, Anime or going back to Pony, and then doing a I2I + ControlNet workflow... OR create a new style LoRA for Flux 2 Klein that would allow me to easily convert with a simple unified prompt in edit mode and with a high-chance of a one-shot generation. I chose the latter.

I am really not familiar with the extensive anime style types, so I was hoping there was an existing F2K9 style LoRA that would fit the bill but I couldn't find one so I cobbled together a dataset with what I was going for, both hybrid generated and broadly grabbed, and cherry-picked those which matched the ascetic I was going for as far as drawing style, facial geometry, body morphology, and colors & contrast.

After I popped the dataset into the AI Toolkit oven, and a little while later, I was very pleased with the results as it did exactly what I wanted: Giving me an easy way to transform a large set of exiting images into digital illustrations with a high-degree of compositional fidelity avoiding the need to individually tag/caption + i2i+CP using an illustration specific checkpoint like Illustrious or Anime.

So, after I was done I thought I'd try it out to generate images and was also very happy with its complex prompt adherence results with subject pose/positioning, and naughty bits. It's not perfect, but to me it's pretty darn good and darn near perfect if you go the extra mile by adding on some CN.

While I'm sure there's anime aficionados that will point out the inferior nature of Flux 2 Klein, compared to other base models, when creating digital art/animation but at least with my novice eyes the t2i generations were comparable to the original dataset and other similar styles based off of anime-focused checkpoint models. So I thought I'd share it in case the community might also find it useful.

The image gallery examples use typical out of the box configurations, e.g. Euler, 4-steps and a plane-Jane Flux 2 K 9B base checkpoint, to show a floor baseline on what type of outputs it can generate out of the box. Your results may exceed them if you use better samplers, higher step counts, custom merged checkpoints, etc. The gallery contains both edit mode transformations and full image generation examples.

I'm afraid I don't know the style-type name or if its based off an individual artist or sub-genre in general, since my focus wasn't to add myself to the hat of the plethora anime-focused LoRAs creators and instead very project specific, but shared anyway in case its even a little bit helpful for others. Hence the model title name, however if you believe you know that it's of a particular anime style classification type then let me know and I can update the description to make it easier for others to find if they search for that.

Activation Note: Not particularly tied to a trigger word but make sure you specify somewhere in your prompt (generate or edit) digital illustration, anime, cartoon, or something similar to let it know your looking for an non-realistic output.

Enjoy.