"Please read this before using, thank you."

Generate simple anime style portraits for Flux Dev.

(Currently Works for Simple Prompts Only!)

Still in BETA phase, expect instabilities.

Why still in Beta and not v1? It's because I'm not confident enough to call this a "good" model for a proper release.

This version has quite a different behavior than Beta 2. Previously when you use natural language, you needed to actually add the trigger in the prompt to make the lora stick but that isn't needed when you use tag prompting. In Beta 3, you don't really need to use the trigger and instead just use higher weights (0.8 is the max I think) and to mix the style into 2.5D-ish, you use natural language + lower weight (0.5 to 0.6). But if you use tag prompting, make sure that you have the weight set to 0.5 (can be lower actually) because the effect is very strong when you use tag prompting.

Changelog (Beta 3):

> Trigger word changed into "anime".

> Improved the hand quality a bit (it still gives you weird hands sometimes but as far as testing goes, it has better hand shape than Beta 2.

> Changes to the training parameter resulting in quite a difference in behavior (compared to Beta 2).

Known issue:

> Feet sucks, for now

> You tell me

====================================================