Knowledge Base
  • Dreamerland
  • workflow
    • Hires. fix
      • Hires. fix with different model
    • Optical illusion
    • Inpainting
      • BrushNet
      • Inpaint not masked
    • Regional Prompt (color mask)
      • Regional Prompt with Reference Image
    • Regional Prompt (prompt mode)
    • Outpainting
    • Image prompt
      • Consistent face
      • Consistent anime character
      • Style
    • Doodle
      • Inpaint with doodle
      • Area inpaint with doodle
    • Motion brush
    • Relighting
    • Live Portrait
    • Face Swap
    • 3D Parallax effect
    • In-context transfer
  • Video
    • Image to Video
    • Hunyuan Video
      • Image to Video
    • Wan2.1 Video
      • Pose reference
      • Video to Video
        • Reference Image
        • Key frames
      • Inpaint
        • Free mode
  • Basic terms
    • Sampling method
    • Guidance scale
    • Negative embeddings
    • CLIP skip
  • FAQ
    • Dreamerland Web app
    • Prompt syntax () and []
    • Bad quality image or deformed image
    • High res. fix vs tiled upscale
    • How to enable sensitive content
    • Wildcards
Powered by GitBook
On this page
  1. Basic terms

Guidance scale

The Guidance Scale Stable Diffusion, or Classifier-Free Guidance Scale (CFG), is used in AI to decide how much an AI art creation follows the text prompt. It's basically a number that shows how important the text or image description is when generating AI art.

If you crank up the guidance scale, the generated images will stick closely to the provided text prompts, but they might lack that spark of creativity. So, it's better to aim for a lower CFG value when creating images, but not too low either. Usually, keeping the numbers between seven and nine will give you high-quality results, unless you really know what you're doing with a high CFG setting.

Last updated 1 year ago