Scalable High-Resolution Pixel-Space Image Synthesis with Hourglass Diffusion Transformers

1 code implementation • 21 Jan 2024 • Katherine Crowson, Stefan Andreas Baumann, Alex Birch, Tanishq Mathew Abraham, Daniel Z. Kaplan, Enrico Shippole

We present the Hourglass Diffusion Transformer (HDiT), an image generative model that exhibits linear scaling with pixel count, supporting training at high-resolution (e. g. $1024 \times 1024$) directly in pixel-space.

Image Generation