Let's chat about VEO2 and The Heist. This is a multi-thread post that will discuss prompts and techniques. As an overview, this test was mainly to see if you could tell and cut together a short story (albeit simple) via text-to-video. Also note, that VEO2 is in beta.

I tried to see how Kling v1.6 would handle the trolley problem. But it just backed away slowly.

Video generation models exploded onto the scene in 2024, sparked by the release of Sora from OpenAI. I wrote a blog post on key techniques that are used in building large video generation models: yenchenlin.me/blog/2025/01/0…

Introducing Ray2, a new frontier in video generative models. Scaled to 10x compute, #Ray2 creates realistic videos with natural and coherent motion, unlocking new freedoms of creative expression and visual storytelling. Available now. Learn more lumalabs.ai/ray.

Introducing Spell, a model to generate 3D worlds. A new era for 3D. ✨ (Thread - Link Below)

Announcing Diffusion Forcing Transformer (DFoT), our new video diffusion algorithm that generates ultra-long videos of 800+ frames. DFoT enables History Guidance, a simple add-on to any existing video diffusion models for a quality boost. Website: boyuan.space/history-guidan… (1/7)

cool!

Today, we release Inductive Moment Matching (IMM): a new pre-training paradigm breaking the algorithmic ceiling of diffusion models. Higher sample quality. 10x more efficient. Single-stage, single network, stable training. Read more: lumalabs.ai/news/imm

🚀🚀🚀Introducing VideoScene (CVPR'25) - a turbo upgrade of ReconX! Our one-step video diffusion model bridges the gap from video to 3D, outpacing slow multi-step pipelines. Paper: arxiv.org/abs/2504.01956 Project Page: hanyang-21.github.io/VideoScene Code: github.com/hanyang-21/Vid…

wow

Introducing Multiverse: the first AI-generated multiplayer game. Multiplayer was the missing piece in AI-generated worlds — now it’s here. Players can interact and shape a shared AI-simulated world, in real-time. Training and research cost < $1.5K. Run it on your own PC. We

A video generator must satisfy 3 criteria to be a world model: 1️⃣ Causality: Past affects future, not vice versa. 2️⃣ Persistence: The world shouldn't change because you looked away. 3️⃣ Constant Speed: Simulation shouldn't slow down over time. We believe SSMs are a natural fit:

Bored of linear recurrent memories (e.g., linear attention) and want a scalable, nonlinear alternative? Our new paper “Test-Time Training Done Right” propose LaCT (Large Chunk Test-Time Training) — a highly efficient, massively scalable nonlinear memory with: 💡 Pure PyTorch

Real-time video generation is finally real — without sacrificing quality. Introducing Self-Forcing, a new paradigm for training autoregressive diffusion models. The key to high quality? Simulate the inference process during training by unrolling transformers with KV caching.

I converted one of my favorite talks I've given over the past year into a blog post. "On the Tradeoffs of SSMs and Transformers" (or: tokens are bullshit) In a few days, we'll release what I believe is the next major advance for architectures.

What exactly is a "world model"? And what limits existing video generation models from being true world models? In my new blog post, I argue that a true video world model must be causal, interactive, persistent, real-time, and physical accurate. xunhuang.me/blogs/world_mo…

thread on the new paper: The Serial Scaling Hypothesis joint work with: @phizaz, Yutong Bai, Kananart

What if you could not only watch a generated video, but explore it too? 🌐 Genie 3 is our groundbreaking world model that creates interactive, playable environments from a single text prompt. From photorealistic landscapes to fantasy realms, the possibilities are endless. 🧵

🚀 My first tweet! (1/n) Thrilled to share our new work: Context-as-Memory (CaM) — tackling the memory problem in Video World Model! Our idea: context=memory. By leveraging context, CaM preserves consistency across generations (like Genie 3). 🎥 Check out our demo video below!

today, we're making another step towards the future. introducing our first Real-time Video generation model. join the beta 👇