PixVid lifts latent video diffusion into pixel space for real-time generation

Roblox, UC Merced and others released PixVid, which lifts a latent video diffusion model into pixel space for frame-by-frame generation without a VAE, doubling resolution per token and enabling real-time, low-latency video generation.

2026-11-14 ~ 2026-11-14 · 2 related posts

• PixVid does real-time video gen frame-by-frame in pixel space, no VAE needed — bdsqlsz · 2026-11-14
• Roblox's PixVid lifts latent video diffusion into pixel space for 2x resolution — Roblox · 2026-11-14