Anima This is my first anima version, this is a merge of multiple loras into preview-3 in different ranks. Its not perfect, it still got problems with hands but its a nice tool for me to generate more data for training with img2img. I think its good enough to share it here for you to use it until i improve it with better data and model training after the full anima release. sampler: er_sde scheduler:simplecfg:4-7steps:20-50 prepending "cartoon" helps with getting more cartoon like results. Illustrious: This is the second version of my checkpoint. It uses a different VAE, aims for a more cartoony style, and differs significantly from v1. To increase the level of cartoon look, use "cartoon" or "toon" in your prompt. Settings CFG 2–6, minimum 15 steps. Euler Ancestral, Karras, or DPM_2, SGM Uniform I prefer CFG 4–6. I used FaceDetailer and lazy embeddings for some of the showcase images. If you use DMD2, the results will look more cartoony. Most of the showcase images were created without DMD2.

Files with SHA256: 401b9caac835fb6f40db5026e1c8385ed8cb9688b49612f5159d1333b5bdf7c3