
This is a semi realism anime style Lora. This is mostly trained on portraits of people. Its got a mix of anime and semi realism images in its dataset. By using the trigger word FFSRA and using a style prompt like "Realistic Anime" you can generate high quality and detailed images. This lora is meant to help keep the images a bit more 3d, a bit more detailed, and make them feel more vivid and intense than you get from running the base Flux model with a similar prompt.
This is not intended to produce realistic results but Flux sometimes overrides the LoRA and makes something a bit more realistic than intended. On the other side if you prompt with solely "Anime" or "Cartoon" instead of "Realistic Anime" you will end up with an image that is a bit more flat but still has more depth and detail than the base model.
This Lora is also meant to allow vivid skin color changes, its not perfect and not all colors work but I've found success by wording it like "Pastel Blue Skin Color" or "Purple Skin Color" its important to use the term "Skin Color" when prompting, it doesn't work well if you leave off "color" when discussing skin colors.
Hands are still an issue in this LoRA. I was picky on the dataset but it appears some issues are still popping up. While its definitely possible to get good hands, it may take a couple of runs to get perfect results on hands.
描述:
训练词语: FFSRA
名称: Flashfire_Semi_Realism_anime_FLUX.safetensors
大小 (KB): 87642
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success