• Merge Here:

• Base XL: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0

• SDXL VAE: https://huggingface.co/stabilityai/sdxl-vae

• The Turbo version is designed to be used at a CFG scale of 2 (3-4 for stylized things) and with around 4-7 sampling steps1. This version should work only with DPM++ SDE Karras (or Normal). It is not recommended to use the LCM1 sampler.

• Additionally, there is no need to use a refiner and can be used for high resolution corrections and tiling1. The examples were generated using Auto1111, but you can get similar results with this workflow from ComfyUI1.

This model is ideal for those looking to generate detailed and realistic images with ease. However, it is always important to remember that the quality of the images generated may vary depending on the complexity and specificity of the instructions provided.