Upscaling Techniques

The process of increasing an image's resolution while maintaining or enhancing quality. AI upscalers can add realistic detail that wasn't in the original.

Upscaling transforms low-resolution AI art into high-resolution images suitable for printing or detailed viewing.

Types of Upscaling

Traditional Upscaling

  • Interpolation (bicubic, Lanczos)
  • Fast but creates blur
  • No new detail added

AI Upscaling

  • Neural networks trained on image pairs
  • Adds plausible detail
  • Can enhance textures and features

SD-based Upscaling

  • Uses diffusion model with img2img
  • Generates new detail
  • Can modify the image significantly
UpscalerTypeBest For
RealESRGANAIGeneral purpose
4x-UltraSharpAIDetailed enhancement
ESRGANAIPhotorealistic
SwinIRAIQuality focused
SD UpscaleDiffusionCreative upscaling

Upscaling Workflow

Basic AI Upscale

  1. Generate at base resolution (512, 1024)
  2. Apply AI upscaler (2x or 4x)
  3. Result: Higher resolution image

Hires Fix (SD)

  1. Generate at low resolution
  2. Upscale with latent or AI upscaler
  3. Run additional denoising passes
  4. Adds more detail than upscaling alone

Tiled Upscaling

  • Process image in overlapping tiles
  • Reduces VRAM requirements
  • Enables very large outputs

Tips

  • Generate at highest practical resolution first
  • Multiple small upscales often better than one large
  • Watch for artifacts in faces and text
  • Combine upscaling with face restoration