This week in Multimodal

Back to Pixels: Lifting Latent Video Diffusion into Pixel SpaceNov 12 · ▲ 3
Unbundling Multimodal Vision EncodingNov 12 · ▲ 13
TTB: Test-time Baking for Efficient Rendering of Decoder-only View Synthesis ModelsNov 12 · ▲ 21

This week

Get the morning digest

The top stories in the fields you follow, at 08:00 your time.

Latest in AI and developer tools

Today's trail

updated 5m ago

Last hour Today

Pulse · marktechpost.com

marktechpost.com · Nov 14

Stanford Research Open-Sources Memorilla, a Latent Memory Module for Frozen LLMs

Pulse · techtimes.com

techtimes.com · Nov 14

Baidu Pledges Apache License for ERNIE 5.5 AI Weights Due This Month; Base Model Bars EU, UK

ERNIE 5.5 open weights face an unsettled license: Baidu’s product page pledges Apache 2.0 for the Preview model, yet its release names no license and sets a Nov. 25 deadline, while the Base model bars the EU, UK and South Korea. Developers also weigh Chinese state-access law and unverified Token

Pulse

Top picks this weekarXiv

arXiv

Nov 12

Back to Pixels: Lifting Latent Video Diffusion into Pixel Space

Frontier video generation models work in a VAE latent space that compresses video in both space and time. This compression made training tractable, but now limits high-resolution, controllable, and interactive generation -- doubling resolution quadruples token count, conditions apply to groups of frames rather than individual ones, and each response must wait for the rest of its multi-frame block. We present PixVid, a simple yet effective method that lifts a pretrained latent video model to directly generate every frame in pixel space. PixVid keeps the diffusion transformer and replaces the VAE decoder with a convolutional head that renders its outputs into pixels. We find that predicting clean video rather than velocity, together with a larger noise shift, is key to quality. Finetuned from Wan2.1-14B, PixVid is competitive with the latent baseline within $\sim 1.3$M training clips. Working framewise in pixels, it generates at $2\times$ higher resolution with the same per-frame token count, keeps depth-conditioned video editing aligned under fast motion, and responds to each action at the next frame. Project page: https://chenshuo20.github.io/PixVid/