Christmas in May — Anima 1.0 is out! c:
TAGS[artist] [style] [color palette] [1girl/1boy/etc] [character] [series] [others]Quality scores: No more positive quality tags..
Negative: score_1, score_2, score_3, worst quality, low quality, blurry, lowres, extra fingers, bad feets, bad eyes, logo, noisy, watermark, signature
The watermark consists of a pair of trigger words and that display (~in the case of this model) a "text" in the middle of the image; this lets me know which model is merged with this one, no matter how many times it's mixed.
✦•··· Recommended Settings ···•✦
Steps: 12 ~ 40 (Civitai generator 20+)
CFG scale: 5 ~ 6
Sampler: EulerA / dpmpp_2m_sde_gpu - SGM_Uniform / Simple/ Beta
Hires-fix: Same settings you used to generate the image..
Description
Fourth Iteration 𐔌՞. .՞𐦯
A couple of minor changes compared to v2.0..
ID watermark on weights (doesn't affect quality)
Goodbye to positive quality tags.
CosmosPredict2.5 DMD2 Low-step inference:
- 12 Steps and CFG 5 or superior..Available Precisions:
- FP16 / NVFP4 / INT8_CONVROT