For best outputs: Samplers: DPM++ 2M, DPM++ 2M SDE Karras, DPM++ 2M Karras, DPM++ 2S a Karras, Euler a. Steps: 20 ~ 80, cfg around 8, CLIP skip 1. VAE: sdxl_vae.safetensors. Style: This model has the ability to create a variety of styles but focuses on 2.5D like images. Kind of generations: Fantasy Sci-fi Anime Semi-realistic Painted Landscapes Psychedelic High Details Artistic Problem areas This model has some difficulty with hands when describing complex prompts, so keep your prompts short. It wasn't trained on any realistic images, so I cannot confirm if it has that capacity, but you can reach certain level of realism by using the right words in your pos, like for example "high-definition photograph" or "HDR". Not trained on any NSFW work, although can create nudes.