Skip to content

StableDelight

Remove specular reflections to reveal hidden texture

Visit official site

What StableDelight is

The model is described as removing specular reflections from textured surfaces to reveal hidden detail, extending the stability idea of StableNormal. It adds a multi-scale SSIM loss and random condition scaling during diffusion training.

What you can do with it

  • Clean product surfaces before defect inspection
  • Remove glare while digitizing artworks
  • Reduce reflections in medical images

Getting started

  • Open a terminal or command prompt
  • Install the package with pip install stable-delight
  • Import torch and the PIL Image class
  • Load an image with Image.open

Pros & cons

✓ What we like

  • One-step diffusion prediction is sharpened
  • The model inherits StableNormal stability

! What to watch out for

  • Training relies on several external datasets
  • Loading the model requires a Torch environment

FAQ

Which model is it based on?

StableNormal is named as the basis for its stable monocular normal estimation.

Is there a ready interface?

A Gradio interface and a Torch Hub Loader are listed.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review