Edit: had some minor problems with V1 with turbo lora so fixed it up with V2. Keeping V1 there though in case ppl still find it better.

also not letting me set trigger words for V2 (whi ch are the same so ill post it here):

long hair, purple eyes, grey hair, pointy ears, braided ponytail, <lora:DSAnima Emilia S4 V1:2>

hood up, animal hood, x hair ornament,

hood down, hair flower, braid, x hair ornament,

white cape, long sleeves, purple bodysuit, white footwear,

Kinda a throwaway lora i figured i would upload. just wanted a lora with her in that outfit because i thought it was cute lol and to test Local Training vs. Civitai AIToolkit training and well, i will be sticking to local, the AI Toolkit is bad for Anima(at least civitai's version) and i recommend training local if possible.

It looks fine on the surface but i find that it has worse prompt adherence and doesnt seem to work well with other loras (particularly concept loras). It also still does that weird thing where it puts the odd filename instead of the trained filename when you add it to your prompt.

I have a ton of datasets i would of loved to throw into civit trainer to get it out to you guys but it looks like the best course of action will just be to drip feed it to you guys by training them overnight / when i am away from home.