사용자 LoRA
This is my 21th Lora! It is not the best but still can generate some good images.TRIGGER: torn clothes, dirty, homeless, on street, garbage, sitting, long hair, outdoors(no animal ears) - Add this to ensure close resemblance to char (optional)Tested on all Illustrious models. I have not tested on any other.Steps: 20-30 (use model recommendation / I trained on 24 steps)Cfg: 7.0 (Use what is recommended by model)Sampler: Euler_AScheduler: normal / simple / beta (may produce artifacts)HIGH RES FIX: ((((highly recommended)))Steps: 13Denoise: 0.58Upscale type: latentLatent upscale: nearest_exactUpscale by: 1.5FOR UPSCALER: 4x RealESRGAN X PLUSDaemon detailer (recommended) {Can be ultra good at 0.55 denoise and 20 steps / I have seen it brings out a lot of detail and also fixes certain issue with some anatomy, do not go to 0.6 as it can change the image a lot, stick to 0.5-0.55}Face and hand detailer (Optional)HAPPY GENERATIONS!