More than 25 million people are using Google Flow every month to dream up new ideas, create stories, and build cool things. Thank you. We’re continuing the 50 additional daily credits for all users. Dive in and keep creating.

流行のJevをMiniMax H3に組み込んで、動画生成を高速化してみた！Attention処理のスパース化にJevを使用。 ・層ごとにJevが重要度を判定（4step 49層が対象） ・Jevがスパース率1%, 3%, 5%, 10%を選択 RTX4070で6分7秒→3分34秒で41.7%短縮！動画生成中にJevクラウドに問合せしているのに速い！

この3x3画像を1枚 r2v でAI動画化すると以下の設定でおおよそそのまま映像になる Minimax H3 Max r2v 480p 15秒　Prompt調整：Quality　参照強度：標準 「左上から右下のパネルにカットが切り替わる2Dアニメーション、 複数パネル禁止、BGM禁止」

Introducing VC-Attention: fast and accurate low-bit attention without retraining. On MiniMax-H3, VC-Attention speeds up attention by 1.6× on B200 and 1.5× on B300 over FlashAttention-4, with better fidelity than SageAttention2. It also works with existing sparse attention methods. Two key innovations: • V-Smooth reduces value quantization error. • ExpCast-FP8 speeds up softmax. Nunchux Attention, our proprietary extension, pushes the speedup to 1.9× on B200 and 1.8× on B300. Blog: http://www.nunchux.ai/blog/attention-is-the-video-bottleneck Technical Report: http://arxiv.org/pdf/2609.15810 Joint work by researchers at MIT, CMU, UC Berkeley, Stanford, and NVIDIA.

I gave Astra one brief and it delivered the whole scene in Blender, camera direction included. All of it ran inside MiniMax Design. Astra sits there as the agent and talks to Blender over an official connector, so everything syncs straight onto the canvas. The final video is MiniMax H3, from the same canvas. #MiniMaxdesign - https://design.minimax.io/

I used the work generated with GPT-6 Astra to build out a set, a story. Now I'm bringing it to life with @Hailuo_AI Minimax H3. Here's a tip that's worked well for me 👇

SGLang-Diffusion with VDN-H3 now generates 14.4s of 768p video in just 9.0s 🚀 On 8× B200, 8 step denoising takes just 6.9s, reaching over 2× real time. The 9.0s figure covers the full generation request after warmup. No measured quality regression versus dense 50-step H3 across 103 test prompts. 🧵

Final call. Bestiary closes tonight, Sep 14 at 23:59 (PT). ⏳ $8,000 cash, 200,000 Credits, and ten Audience Choice awards are still on the table — and they go to the people who actually hit submit. One strange, beautiful short film, 30 seconds or longer, generated with MiniMax H3. Any myth, any era, any world you can dream up. The bestiary doesn't close itself. Finish your film before the clock runs out.

🚀 What if video generators could build on representations that already understand the visual world? We are excited to introduce V-RAE: Rethinking Video Latent Spaces for Generation. Recent progress in image generation has begun to move beyond conventional VAE latents, exploring both direct pixel-space and representation-based approaches. Video generation, however, still depends heavily on latent compression, as the scale and redundancy of spatiotemporal data make direct modeling prohibitively expensive. However, most video VAEs are optimized for pixel reconstruction, and a latent space that reconstructs well is not necessarily easy to generate. V-RAE takes a different approach: it directly uses representations from frozen vision foundation models as the generative latent space, rather than as auxiliary supervision. We study DINOv3, SigLIP2, EUPE, and V-JEPA 2.1. A lightweight temporal attention pooling module compresses their dense features by 4×, followed by a spatiotemporal Transformer decoder. Under matched generation backbones, latent budgets, and training settings, V-RAE achieves: 🏆 2.13 rFVD on Kinetics-600 🎬 117.86 gFVD on UCF101 and 19.16 gFVD on Kinetics-600 ⚡ Up to 6× faster convergence than VAE-based latent spaces 🧠 90.92% semantic probing accuracy on UCF101 🌍 Better future prediction on Cityscapes, reducing gFVD from 144.47 to 111.36 Our experiments also reveal a broader finding: Good Reconstruction ≠ Good Generation. During generation, predicted latents inevitably deviate from real encoding trajectories. If the latent space is not sufficiently smooth, small errors can be amplified into visible artifacts. We therefore introduce tFVD to evaluate temporal smoothness and robustness to latent prediction errors. It correlates much more strongly with downstream generation quality, reaching 0.919 on Kinetics-600. The takeaway: A latent space is not merely where videos are compressed—it determines what the generator must learn. When semantics and temporal structure are already organized in the representation, generation becomes easier to learn. Representation first. Generation follows. Many thanks to my mentors, @ScottNLP and @SQWu_Tori, for their continuous guidance and support. I am also deeply grateful to @sainingxie for his valuable guidance and invaluable feedback, which greatly helped shape V-RAE. 🙏 Hi @_akhaliq, we would truly appreciate your help in sharing V-RAE with the broader AI research community. Thank you! 🙏 📄 Paper: https://arxiv.org/abs/2608.13556 💻 Code: https://github.com/V-RAE/V-RAE 🤗 Models: https://huggingface.co/Guomh0707/V-RAE-Models 🌐 Project: https://v-rae.github.io #VideoGeneration #GenerativeAI #ComputerVision #WorldModels #RepresentationLearning #RAE

Can MLLMs actually track what's happening in a video? Introducing VSTAT 🎯, our new benchmark for visual state tracking. The tasks are simple: count cups, read typed words, count page flips. Humans solve them easily. MLLMs don't. https://vision-x-nyu.github.io/vstat-site/ 🧵 [1/11]