Nvidia just dropped Describe Anything on Hugging Face Detailed Localized Image and Video Captioning

We released Cosmos-Reason1 code, model, and part of the data! We also updated our paper to include a section about our RL infra: arxiv.org/abs/2503.15558 - Code: github.com/nvidia-cosmos/… - Model and Data: huggingface.co/collections/nv… - Blog: developer.nvidia.com/blog/curating-…

1/💡New paper from NVIDIA&Tsinghua International Conference on Minority Languages Spotlight! Direct Discriminative Optimization (DDO) enables GAN-style finetuning of diffusion/autoregressive models without extra networks. SOTA achieved on ImageNet-512! Website: research.nvidia.com/labs/dir/ddo/ Code: github.com/NVlabs/DDO

Check out our new work on Direct Discriminative Optimization improving GenAI models.

Cosmos-Reason1 has exciting updates 💡 Now it understands physical reality — judging videos as real or fake! Check out the resources👇 Paper: arxiv.org/abs/2503.15558 Huggingface: huggingface.co/nvidia/Cosmos-… Code: github.com/nvidia-cosmos/… Project page: research.nvidia.com/labs/dir/cosmo… (1/n)

We post-trained a reasoning model to reason whether a video is real or generated. It might be very useful as a critic to improve video generators. Take a look. NVIDIA AI

For people looking for a diffusion-based video generator to finetune or post-train for their downstream physical AI applications, we just released our latest one. We have 2 models: 2B and 14B. 2B for fast prototyping and 14B for better quality. The license is fully open. Give it

Happy to share our work PartPacker: We enable one-shot image-to-3D generation with any number of parts! Project page: research.nvidia.com/labs/dir/partp… Demo: huggingface.co/spaces/nvidia/… Code: github.com/NVlabs/PartPac…

Generating 3D models with parts is a key step toward scalable, interactive simulation environments. Check out our work — PartPacker — and the concurrent project, PartCrafter!" PartPacker: github.com/NVlabs/PartPac… PartCrafter: wgsxm.github.io/projects/partc…

3D asset generation has advanced a lot in the past few years. Generating a holistic 3D asset is no longer a challenging problem. What's next for 3D generation? We believe that generating a 3D asset with individual parts defined is the next frontier. With the parts, we can start

Check out our latest HF demo on 3D generation with part annotation.

Big congrats to Eric Jang and the team on the 1X World Model release. Verification is an important part of producing production AI model. Given the diverse nature of the work environment, it makes a lot of sense to leverage a world model to help with policy evaluation.

We build Cosmos-Predict2 as a world foundation model for Physical AI builders — fully open and adaptable. Post-train it for specialized tasks or different output types. Available in multiple sizes, resolutions, and frame rates. 📷 Watch the repo walkthrough

🤖🌎 We are organizing a workshop on Robotics World Modeling at Conference on Robot Learning 2025! We have an excellent group of speakers and panelists, and are inviting you to submit your papers with a July 13 deadline. Website: robot-world-modeling.github.io

We at 1X with Jack Monas are excited to announce the ICCV phase of our 1X World Model Challenge: huggingface.co/spaces/1x-tech… Participate in the Compression and Sampling tracks for a $8k prize pool & train generative models for cool robot results like: 1x.tech/discover/redwo…

Together with Aaron Lefohn and Sanja Fidler, we will give a special address at SIGGRAPH. Specifically, I will give an update on our vision and our current work in enabling Physical AI. Please join us. nvidia.com/en-us/events/s…

Kick off your #OpenUSD Day with a look into the future of robotics and autonomous vehicles. 🤖 Join Ming-Yu Liu as he shares how #NVIDIACosmos world foundation models unlock prediction and reasoning for the next wave of robotics and autonomous vehicles. 📅Wednesday, 8/13 at

In Cosmos, we are hiring Cosmos World Foundation Model builders. If you are interestd in building large-scale video foundaiton model and multimodal LLM for Robots and cars, please send your CV to [email protected] If you have experiences in large-scale diffusion models,

The submissions portal for the NVIDIA 2026-2027 Graduate Fellowships is now open research.nvidia.com/graduate-fello… PHD students work on AI. Please apply!

[1/N] 🎥 We've made available a powerful spatial AI tool named ViPE: Video Pose Engine, to recover camera motion, intrinsics, and dense metric depth from casual videos! Running at 3–5 FPS, ViPE handles cinematic shots, dashcams, and even 360° panoramas. 🔗 research.nvidia.com/labs/toronto-a…