Here is Prediction Boost implementation as stable-diffusion-webui-forge extension.
You can generate more bright and clear images with this.
Currently, only stable-diffusion-webui-forge is supported.
Your feedback is welcome.
Install
Extensions -> Install from URL -> URL for extension's git repository
Put this URL: https://github.com/tmagara/sd-forge-prediction-boost
Settings -> Reload UI
Example
a witchcraft with rounded bob hair and vivid eyes, sophisticated style, actual location
Negative prompt: unknown author
Steps: 15, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 42, Size: 1024x1024, Model hash: 31e35c80fc, Model: sd_xl_base_1.0, VAE hash: 235745af8d, VAE: sdxl.vae.safetensors, Version: f0.0.17v1.8.0rc-latest-276-g29be1da7
Boost scale = 0.0
Boost scale = 0.05
Boost scale = 0.10