QingHe-SDXL-Fantasy Engine

V2.0- Darkroom Image Generation:
Version 2.0 incorporates a large number of realistic photography materials, blending many cinematic stills styles, fixing some body parts, enhancing the realistic texture of models. The current model's style is not fixed, supporting both realistic and anime-style, with models labeled using natural language, it's recommended to use natural language for prompt words.

【Recommended Parameters】:(Reading through assists in better use of the model)

Realistic Quality Words:(If you want to draw realistic characters, it is recommended to use these for better texture)

((film still, movie still)), photograph, volumetric lighting, Slow Shutter Speed, Primitivism,
RAW photo, ultra-sharp focus, internationally acclaimed photograph, ambient atmosphere, ambient,

Positive Quality Word References:

  (masterpiece, best quality, ultra-detailed, absurdres)

Negative Prompt Word References:

  (worst quality: 1.4), (low quality: 1.4), bad eyes, airbrushed, plastic, deformed, watermark

Sampling Method (sampler): DPM++ 2M Karras(Different samplers have different effects, you can try to find one that suits you),  

Iterations Steps: 30

CFG Scale: 7    

Resolution Reference: Recommended to use 832*1216      If the device allows, you can enable 2x HD repair to prevent facial blurring.

If your device configuration is limited like mine, in this resolution.strongly recommend enabling AD facial repair to prevent facial damage., Avoid facial collapse.