This model is an accelerated version of the HelloWorld SDXL base model, incorporating two technologies from SDXL-Lightning. Equipped with the Euler sampler and CFG 1, it can generate images in 6–8 steps, three times faster than the original SDXL. Furthermore, compared to LCM or Turbo versions, its image quality is even more outstanding.

GPT-4o was used to re-caption all datasets. A structured captioning method was adopted this time, with the specific structure being: "One-sentence summary description + multiple image element tags + inspired by XXX + aesthetic quality descriptors". The aesthetic quality descriptors are divided into five levels: worst quality, low quality, medium quality, best quality, and masterpiece. A typical captioning example is as follows:

conceptual art featuring a human hand wrapped in red and beige ribbons, isolated against a plain, light background, realistic style, minimalist color scheme, smooth textures, elongated and surreal aesthetic, inspired by salvador dalí's surrealist works, masterpiece

Recommended parameters for generating images with this model: Sampler: Euler a (Important! This model is specifically tuned for Euler a; other samplers may not yield good results) CFG scale: 1 Sampling steps: 8 steps (6–8 steps both work) Hires upscaler: ESRGAN 4x / 8x_NMKD-Faces_160000_G Hires Upscale factor: 1.5x Hires steps: 8 steps Hires Denoising strength: 0.3