Skip to content

Diffusion-RWKV

A scalable RWKV-based diffusion model for images

Visit official site

What Diffusion-RWKV is

The model is described as a diffusion model based on the RWKV architecture, designed to improve the scalability of diffusion models for image generation.

What you can do with it

  • Generate images without a class condition
  • Generate images conditioned on a class
  • Use the scalable RWKV architecture

Application scenarios

  • Train on ImageNet for 256x256 high-quality images
  • Train on CelebA for 32x32 unconditional images
  • Produce class-specific images such as animals or vehicles

Pros & cons

✓ What we like

  • Optimised for image generation tasks
  • Both training modes are supported in one model

! What to watch out for

  • The named results are tied to specific datasets
  • Output size depends on the model variant and dataset

FAQ

What training modes are listed?

Unconditional and class-conditional training are stated.

Which model variants are named?

DRWKV-H/2 and DRWKV-B/2 are listed.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review