Du hast einen ComfyUI-Workflow? Erstelle direkt im Browser eine verständliche Workflow-Dokumentation.Workflow verständlich machen.Jetzt ausprobieren ↗
thank you mr astromahdi from gpu.garden
Modellquelle
Quellenbeschreibung
thank you mr astromahdi from gpu.garden
Please support the original artist: GumiaoAndPeace
(Anima 1.0) This LoRA is the first to be using a better training config that I recently received from someone else, for this lora at least it seems to be less overfit on composition and seed while still having similar level of style accuracy to my old config. It also train x1.5 time faster so lol W for me.
(Preview 3) All preview images were genned with V2. I feel like the lora is definitely not accurate enough, it didnt really capture the full texture and scratchyness feel of the style, but I dont think I could push Anima to learn further without harming stuffs so oh well. Hope Anima full version style convergence will be even better.
Please support the original artist:
Quellen
1 QuelleModellartefakte
1 Artefaktgumiaoandpeace_LoCon-AnimaBaseV1.0.safetensors-Ryusho-config-000013.safetensors
safetensors · BF16 · 68,9 MB · SHA-256 3ca9d1d8a068…c2cd · Civitai
HerunterladenQuellenauszüge
1 AuszugSource context: 200 downloads · Type LoCon · Base model Anima
V1: Trained on a bigger dataset, but I didn't found it to be as accurate, but it could be more stable and less rigid.
V2: Narrowed dataset down to 25 image and abuse repeat (OP strat), seems a lil more accurate, but may be less stable.
Disclaimer
I respect the artists who honed their crafts for years to achieve these styles among all else, if any artists whom LoRA i've made from shown displeasure/objection to their arts being used as AI training material or if they request a takedown notice on Danbooru , I'll delete the LoRA with no further notice.
