SDXL with transparency (Layer Diffuse)

Write prompts in English, it's SDXL without neuro-assistants. Generation is most stable at 1024x1024 resolution.