
Default level of detail is semi-realistic. However, it is not limited to this art style. You can prompt for anime or hentai or pixiv or other similar prompts to draw something else. For landscape, I've found if you prompt for "landscape-photography" it will try to draw without people unless you prompt for it. You don't have to prompt for "photorealistic" but you may want to negative prompt it to achieve other styles.
This model is a merge between my base model "ACR" + a model I've dubbed "ACR_sparkle" (not uploaded to civitAI) with select merges.
Sparkle consists of several iterations of models with ACR including:
NeverEnding Dream (NED) - v1.22 baked vae | Stable Diffusion Checkpoint | Civitai
Komix 疯丹-翠眸烈舞 - v1.0 | Stable Diffusion Checkpoint | Civitai
hellophotobox - hellophotobox_v1.1c | Stable Diffusion Checkpoint | Civitai
Notes about sparkle (and I may upload this version later): I noticed several improvements with Sparkle over ACR which included better fingers naturally on regular images without prompting for it, or using any Loras.
Fusion consists of Sparkle +
hellophotobox - hellophotobox_v1.1c | Stable Diffusion Checkpoint | Civitai
Synergix - v1.0 | Stable Diffusion Checkpoint | Civitai
-
Deliberate_v2 (I couldn't find it listed on CivitAI, but this version was downloaded last July.)
Notes about Fusion: On the whole, fingers drawn are better than ACR without prompts or Loras.
I "interviewed" over 100 models in this selection to come up with the best combinations with the least errors. Hope you enjoy using this model!
描述:
More emphasis on anime style with a realistic edge.
训练词语:
名称: acrFusionFloat16_anime.safetensors
大小 (KB): 3102686
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success