SVG-T2I: Scaling Up Text-to-Image Latent Diffusion Model Without Variational Autoencoder
IntermediateMinglei Shi, Haolin Wang et al.Dec 12arXiv
This paper shows you can train a big text-to-image diffusion model directly on the features of a vision foundation model (like DINOv3) without using a VAE.
#text-to-image#diffusion transformer#flow matching