Skip to content

X-Adapter

Let pre-trained plugins work with upgraded diffusion models

Visit official site

What X-Adapter is

The tool is described as a general upgrade adapter that lets pre-trained plugins work with an upgraded text-to-image diffusion model without further retraining.

What you can do with it

  • Make pre-trained plugins compatible with upgraded models
  • Keep the connectors of the old model
  • Add trainable mapping layers for feature remapping
  • Apply empty-text training and two-stage denoising

Application scenarios

  • Use X-Adapter with an upgraded SDXL model
  • Apply pre-trained SD 1.5 plugins to an SDXL model
  • Demonstrate it across different plugins and base models

Pros & cons

✓ What we like

  • No further retraining of the plugin is stated
  • The old model connectors are kept

! What to watch out for

  • An extra network has to be trained
  • The setup targets specific model versions

FAQ

Which plugins are named?

The material names ControlNet and LoRA as examples of pre-trained plugins.

Does it need retraining?

The material states plugins work with the upgraded model without further retraining.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review