Who wants Open Source Veo 2?

📢📢📢 Consider applying to SFU, wher we have one of the largest Graphics/Vision groups in the world. Still in time! Deadline is Jan 18, 2025 The language of Vancouver is English, and despite being in 🇨🇦... it is not that cold (similar to Paris, Berlin, London). Links in 🧵

The future. This is Luma AI #Ray2 💕🥰

The wait is over… The new SOTA in the video generation is here. This is Luma AI #Ray2

Preprint of (not) today: Liang et al., "Feed-Forward Bullet-Time Reconstruction of Dynamic Scenes from Monocular Videos" -- research.nvidia.com/labs/toronto-a… A ViT that takes multiple frames (and their time andpose in plücker representation) outputs 3D Gaussians at queried time.

🚀 Introducing Pippo – our diffusion transformer pre-trained on 3B Human Images and post-trained with 400M high-res studio images! ✨Pippo can generate 1K resolution turnaround video from a single iPhone photo! 🧵👀 Full deep dive thread coming up next!

We're hiring our first product designer at World Labs -- come help us explore how AI could reshape how we create and interact with 3D spatial content! Seeking candidates w/ 5+ years of experience and ability to work in-person in SF. DMs open.

In the past 1.5 weeks, there appeared 2 papers by 2 different research groups which develop the exactly same (and embarrassingly simple) trick to improve convergence of image/video diffusion models by 20-100+% (sic!) arxiv.org/abs/2502.14831 arxiv.org/abs/2502.09509

📹 EventSplat: 3D Gaussian Splatting from Moving Event Cameras for Real-time Rendering Toshiya Yura, Ashkan Mirzaei, Igor Gilitschenski 5/🧵 arxiv.org/abs/2412.07293

🚀 Have a strong background in 3D/4D Generative Models? Consider applying for an internship with us at Snap’s Creative Vision team! 🎨✨ snap.submittable.com/submit

📢 Introducing a geometry-aware diffusion model for 3D scene inpainting! TL;DR: We condition a diffusion inpainter on a set of coordinate-aligned reference-based cues derived from the scene geometry to autoregressively inpaint a set of images. 🖼️✨ (1/7)

I think this paper represents a pretty significant leap forward for 3D generation. Bolt3D not only generates realistic and renderable 3D scenes in seconds from just a few images, but also our reliance on diffusion means that *completely unobserved* scene regions get generated.

Spatial AI is increasingly important, and the newest papers from #NVIDIAResearch, 3DGRT and 3DGUT, represent significant advancements in enabling researchers and developers to explore and innovate with 3D Gaussian Splatting techniques. 💎 3DGRT (Gaussian Ray Tracing) ➡️

From “Static Concepts” 🖼️ to “Dynamic Concepts” 💃 Super excited about this new work that opens up new challenges in video personalization, and I can’t wait to see the follow ups!

Just wrapped up the first full draft of my 🆕 Gaussian Splatting lecture 😅 Thinking of adding a voiceover … tempting 🤔

Today is one of the most exciting days of my academic career: my first PhD student graduated. 🎉 Huge congrats, Ashkan Mirzaei! 🥂 I am grateful for the opportunity to have worked with you and could not have wished for a better start. 🚀 1/4

image ⇒ video ⇒ 3D/4D I'm super excited to build the next generation of models that understand and can imagine the world like we do at SpAItial with amazing people. Sounds fun? We are hiring! spaitial.ai

In Germany, there is a tradition of creating funny hats for doctoral graduates. 🎓 Claas Voelcker brought this tradition to my group and, together with Umangi Jain, spearheaded the construction of a masterpiece for our first PhD graduate, Ashkan Mirzaei. 1/2

📢 Introducing DenseDPO: Fine-Grained Temporal Preference Optimization for Video Diffusion Models Compared to vanilla DPO, we improve paired data construction and preference label granularity, leading to better visual quality and motion strength with only 1/3 of the data. 🧵

Thrilled to share the papers that our lab will present at #CVPR2025. Learn more in this thread 🧵 and meet Kai He, Yash Kant @CVPR25, Ziyi Wu @ CVPR'25, and our previous visitor Toshiya Yura in Nashville! 1/n