“In contrast to pixel-based ADD, LADD utilizes generative features from pretrained latent diffusion models. This approach simplifies train- ing and enhances performance, enabling high-resolution multi-aspect ratio image synthesis. We apply LADD to Stable Diffusion 3 (8B) to obtain SD3-Turbo, a fast model that matches the performance of state-of-the-art text-to-image generators using only four unguided sampling steps.”
Basically they get almost all of the quality of stable diffusion 3 in 4 steps instead of 50 steps. Seems like a major efficiency win.