Kangfu Mei

I think people don't realize why Gemini Omni is different than other video AIs. It is fully multimodal, so it can edit video natively, too I took the famous "train " movie from 1896 & made it a bullet train, LEGO, added a time traveler, a centipede, muppets... (see reflections?)

@vishalm_patel congratulations!

Big day for the VIU Lab at Johns Hopkins! 🎓 We had an amazing batch of 7 PhD students get hooded today. Huge congratulations to our newest PhDs — proud of all your hard work, perseverance, and accomplishments. Excited to see all the great things you will achieve in industry, research labs, and beyond! #PhD #JHU #AI #ComputerVision @HopkinsDSAI @HopkinsEngineer @JHUECE

Let's rock! 🎸🎸🎸

@CeyuanY Congratulations Ceyuan! That’s great work!

Introducing Omni, one unified model can support any-to-any multimodal modeling, including multimodal understanding, image/video generation and editing, world modeling and 3D reconstruction. All in one that adopts standard mixture-of-experts arch with only 3B activations.

Congratulations Yuyang!

@EvelynZ5699647 This is a great work. Any plan of applying this into video generative model as RL rewards?

🚀🚀🚀We're excited to introduce Galileo 0 (lnkd.in/gJttjEr5) — our first research preview of a world critic for AI-generated video, which already outperforms Qwen 3.5-Plus, Gemini 3.1 Pro, Pegasus 1.2, and GPT 5.4 on physical consistency reasoning 🚀🚀🚀 Galileo doesn't just score outputs. It diagnoses them — identifying what failed, when it failed, where it happened, and why it broke the rules of the world. This is a step toward a new paradigm: generate → critique → refine → repeat — where models don't just produce worlds, but learn to keep them consistent over time. 𝐖𝐡𝐚𝐭 𝐦𝐚𝐤𝐞𝐬 𝐭𝐡𝐢𝐬 𝐦𝐢𝐥𝐞𝐬𝐭𝐨𝐧𝐞 𝐞𝐯𝐞𝐧 𝐦𝐨𝐫𝐞 𝐦𝐞𝐚𝐧𝐢𝐧𝐠𝐟𝐮𝐥: We built Galileo 0 — along with our datasets (including our public Physion-Eval benchmark), evaluation pipeline, and early pilots — with less than $200K total spend in 3 months. No massive training clusters. No billion-dollar budgets. Just a small, relentless team, strong conviction — and yes, at one point, a five-day stretch of not showering to get this model out. Because we believe reliability will become core infrastructure for world models. In a world where billions are being poured into generation, the missing piece isn't more pixels — it's better critics 😊 #PhysionLabs #Galileo0 #WorldModels

Video’s here to stay - introducing Veo 3.1 Lite, our most cost efficient video generation model to date, and on April 7th we are also reducing the price for Veo 3.1 Fast : )

Ten years ago, AlphaGo’s legendary match in Seoul heralded the start of the modern era in AI. Its famous ‘Move 37’ signaled to us that AI techniques were ready to tackle real-world problems in areas like science - and ideas inspired by these methods are critical to building AGI

@baaadas @LumaLabsAI Great job 👏 congratulations

Excited to introduce Uni-1, our new *unified* multimodal model that does both understanding and generation: lumalabs.ai/uni-1 TLDR: I think Uni-1 @LumaLabsAI is > GPT Image 1.5 in many cases, and toe-to-toe with Nano Banana Pro/2. (showcase below)

Congratulations to the team! It is just the first month of 2026.

🧞‍♂️

Happy to share that "Streaming Autoregressive Video Generation via Diagonal Distillation" is accepted by #ICLR26 ! See you all in Brazil 🇧🇷

@CuiMao 这效果也太好了🤣

大家好，这是我第4个fine-tuning的模型版本，以特朗普在达沃斯论坛的发言为主题做了重要播报 这个版本模型更新如下： 1：新增年轻美丽的CuiMao记者lora模型 2：新增CMTV片头节目包装 3：优化TTS效果播音员情感训练输出结果更加稳定 此次CuiMao同志模型的加入意味着CMTV完成了布局的最后一块拼图，下一个版本就是正式版，谢谢大家的支持。

Everyone knows pretraining is powerful, but its potential in IR tasks beyond SFT or RL remains largely unexplored. In this work, we show that reusing pretraining (e.g., Flux) is easier than you think! 🧱 LEGO doesn't need pair-wise data and converts pretrained knowledge in an unsupervised (pseudo-supervised) way. Huge congrats to our intern @YuyangHu_666 for this great work! This is my final paper of 2025. Can't wait to show you what we have for an even more exciting 2026! 🚀

Five generations of our team's interns, all doing so great in their careers. Proud of @_vztu , @KangfuM , @YuyangHu_666 , @hyungjin_chung , and @giannis_daras

Big congratulations 🎉🎊🍾

Great work Yiqun!

Hello from #NeurIPS2025 👋 We're hosting a range of sessions at the @Google booth, including a Q&A with @JeffDean and the Gemini team, plus demos like SIMA 2 – our AI agent for 3D virtual worlds. See the full schedule → research.google/conferences-an…

Interested in Veo 3.1 ? 🔬🔬🔬 From the Veo team @GoogleDeepMind , Sarah and I are presenting a live demo and Q&A at NeurIPS this year! Catch us at the Google Booth on Tues, Dec 2, from 2:00 - 2:30 PM. #NeurIPS2025 #Google

@viddivj Can I slip into the OpenAI's NeurIPS25 party ?🤡

I’m heading to NeurIPS in San Diego this year. Fun full-circle moment: back in 2022 I waited in the long lines and managed to slip into the OpenAI party with my grad school friends — and now I’m attending as part of OpenAI team!