Pinned post
Let's remove VAEs and ViTs from video models!
🚀 Introducing 𝗣𝗶𝘅𝗲𝗹𝗨𝗠𝗠: an encoder-free unified multimodal model for image and video understanding and generation, directly in pixel space.
Code and model available today!
🌐 nv-tlabs.github.io/PixelUMM
📄 arxiv.org/abs/2609.38597
🚀 Introducing 𝗣𝗶𝘅𝗲𝗹𝗨𝗠𝗠: an encoder-free unified multimodal model for image and video understanding and generation, directly in pixel space.
Code and model available today!
🌐 nv-tlabs.github.io/PixelUMM
📄 arxiv.org/abs/2609.38597
0:29