For best outputs: Version 2.0: Sampler: DPM++ 3M SDE Karras. Steps: 70, cfg 7, CLIP skip 2. VAE: none required. Pos prompt: model may tend to create twins. Add "(1girl:1.35), (solo:1.35)" at the beginning of your pos. Neg prompt: "(more than 2 girls:2), (twins:2), (2 people:2), (pieced output), (latex, leather), (worst quality:1.3), (low quality:1), (normal quality:1.4), lowres,skin spots, acnes, skin blemishes, age spot, glan, extra fingers, fewer fingers, strange fingers, bad hand, bad anatomy, fused fingers, missing leg, mutated hand, malformed limbs, missing feet,multiple legs,logo, multiple arms, loli, 3d, extra hands,extra foots, extra legs, ng_deepnegative_v1_75t, ((monochrome)), ((grayscale)), (apron:1.4)", being the embedding "ng_deepnegative_v1_75t" included in my v2.0 example runs. Version 1.0: Sampler: DPM++ 2M Karras. Steps: 40, cfg 7, CLIP skip 2. VAE: none required. Pos prompt: add the LoRa "add_detail" with a weight of 1 at the end of your pos (included in my v1.0 example runs). Neg prompt: "(worst quality, low quality, cgi, bad eye, worst eye, illustration, cartoon), deformed, distorted, disfigured, poorly drawn, bad anatomy, wrong anatomy, (normal quality:2), lowres, bad hands, normal quality, ((monochrome)), ((grayscale)), EasyNegativeV2, paintings, sketches, bad-hands-5", being the embeddings "EasyNegativeV2" and "bad-hands-5" included in my v1.0 example runs. What changes from v1.0 to v2.0 Trinity: input_image feature. Checkpoint Trained.