Skip to content

MuLan

Adapt diffusion models to over 110 languages

Visit official site

What MuLan is

The model is described as an open-source multilingual diffusion model that supports more than 110 languages without extra training, using adaptation techniques for new language settings.

What you can do with it

  • Research multilingual image generation with MuLan
  • Deploy an image generation app for many languages
  • Add multilingual generation services to a product

Model notes

  • Optimise memory use for running efficiency
  • Support base models such as Stable Diffusion XL
  • Support downstream models such as ControlNet and LoRA

Getting started

  • Visit the GitHub page for news and download links
  • Read USAGE.md for install and usage details
  • Pick a base or downstream model and try the Gradio demo

Pros & cons

✓ What we like

  • Base models such as Stable Diffusion 1.5, 2.1, XL and Pixart-Alpha/Sigma
  • Downstream models such as ControlNet, LCM and LoRA

! What to watch out for

  • Adaptation targets a chosen base or downstream model
  • A Gradio demo is provided for a quick trial

FAQ

How many languages are supported?

The material states adaptation for more than 110 languages.

Which base models are named?

The material lists Stable Diffusion 1.5, 2.1, XL and Pixart-Alpha/Sigma.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review