Multi-Scale Pixel-Space Diffusion Transformers Proposed as Alternative to Latent Diffusion
A new arXiv paper introduces multi-scale pixel-space diffusion transformers, treating scale as a modality to avoid the lossy compression and spatial granularity limits of pretrained autoencoders used in latent diffusion models.
Read the full story at arXiv cs.CV ↗
Timeline · 1 report
- 2026-11-10 05:00 · arXiv cs.CV
Scale as a Modality: Multi-Scale Pixel-Space Diffusion Transformers