Hyper3D | Rodin

Excited to share that cuNRTO, our GPU-accelerated framework for nonlinear robust trajectory optimization, will be presented at RSS 2026. 🤖 Robust trajectory optimization is critical for enabling robots to operate safely under uncertainty, but solving the resulting second-order cone programs can be computationally expensive, especially for high-dimensional systems and dense safety constraints. With cuNRTO, we accelerate nonlinear robust trajectory optimization by exposing parallelism in SOC projections and restructuring the inner optimization loop for GPU execution. We evaluate the framework on unicycle, quadcopter, and Franka manipulator tasks, achieving up to 139.6× wall-clock speedup while maintaining robust constraint satisfaction. This project was a true team effort and would not have been possible without the guidance and many insightful discussions with Arshiya Taj Abdul and Prof. Evangelos Theodorou. I'm very grateful for the opportunity to work together across Georgia Institute of Technology, UC San Diego, and Hyper3D | Rodin in this direction. It's great to be back at RSS Conference for a third year, and I'm looking forward to sharing cuNRTO and connecting in Sydney! 💻 Website: https://cunrto.github.io 📄 Paper: https://lnkd.in/gFb4UXYS #Robotics #CUDA #GPUComputing #MotionPlanning #RobotLearning #RSS2026

Last night, I attended fal’s 3D GenMedia Meetup. I really enjoyed the energy, the conversations, and the group of top minds they brought together across 3D and generative media. Thank you Gavin Ni and Tina Sang for the invitation. Jonathan Fischoff shared some thoughtful perspectives on the future of 3D that strongly resonated with me: - Enterprise adoption of 3D generation is becoming real, but the industry is still working toward animation-ready topology, production-ready detail and materials, and project-specific control. This is exactly where Hyper3D | Rodin Gen-2.5 is pushing forward, with controllability as our core philosophy from day one. - The next gaps to close are consistent 3D worlds, believable physics, and stronger creative control. - Reconstruction is also becoming generative. We introduced generative rendering last year, and will be sharing more on this direction soon 👀. One highlight was Benjamin Portman’s demo of PATINA, fal’s new texture model, together with a rendering pipeline using Hyper3D | Rodin Gen-2.5 for base model generation in Blender. The real-time result was amazing. Matt Workman's image blaster and Blendi Bylygbashi's voxel production demo were also top-notch showcases of fal’s creative pipeline capabilities. Ian Curtis’s World Labs Marble showcase was truly impressive as well. We had a great conversation afterward and found a lot of resonance in the product journeys we are both navigating 😄. Hope your hand recover soon! Also great to run into Bhaskara Chaitanya Nandagiri again and hear about his game release using Rodin. Loved the conversations with Dima Dobrynin (great interactive showcase too!), Tanmay Sutar, Jason Kafashi, Ph.D., Qi Guo, YJ Gahng, Andrew Cunningham, Max Wong and Yizhan He. Excited for what’s next in #3D generation and generative media.

We interviewed the CTO of Hyper3D | Rodin to learn all about the latest Gen-2.5 version of #Rodin, which includes controllable #3D #AI workflows and Smart Low-Poly topology generation. Full interview: https://lnkd.in/g-nA477f

Last Saturday in SF, right before #Google I/O, we hosted a room full of top Creative AI builders, partners, and investors, and the conversation went much deeper than a typical public panel. We got into the questions everyone is quietly thinking about: - How do you do GTM when foundation models are converging every quarter? - What does a real ecosystem partnership look like, beyond a logo on a slide? - How do tool companies grow into platforms before big tech absorbs the surface area? A few themes that came up across the room: • The room split on model quality: frontier models still create categories, but workflow and distribution drive repeat usage once quality converges. • Growth paths varied widely: consumer viral, creator-led PLG, enterprise, and infrastructure. Each path shapes product, pricing, team, and fundraising. • The best partnerships were not just co-marketing moments. They were embedded into real workflows. • Big tech is unavoidable, but the smarter path is sharper workflows, communities, and use cases where startups can move faster. Big shoutout to Yuexiang Wang, Lillian Ma, Rebecca Xu and Vera Lin for sharing the unvarnished version of building the fastest-growing companies in this space. I also really enjoyed the conversations with Farid Khasiyev, Arjun Rai, David Zeng, Minh Do, Patrina FAN, Richard Liu, Michael Leung, Michelle Wang, Lyn Zhang, Heroza Z., Tianlong Wang, louis gong, Joshua S. and Bolbi L. And a big thank you to Yuqi Hou, Ruby Xiong, Lynn Duan, niuniu zhao, David Chi, Ali Hassan and Dylan Miao for the generous behind-the-scenes support. Hope we get to meet at the next event! Thanks to PixVerse, GMI Cloud, Lovart AI, VSDesign, AI Plus Founders Community, and Bond AI for co-organizing with us at Hyper3D | Rodin. Hyper3D | Rodin is building the most controllable text/image-to-3D AI, used by millions of creators and companies like NVIDIA, NetEase Games, and more. If you're building, investing in, or helping shape the #CreativeAI ecosystem, would love to connect.

🔥Seedance 2.0 is LIVE on Hyper3D | Rodin FREE UNLIMITED Usage to Business Subscribers! - text to video - image to video - video transition I made this video myself within 1 hr on hyper3d.ai Try it now on https://hyper3d.ai/video

Prowered by Hyper3D Rodin💪 Our developer just won 1st place 🏅 at OpenAI Codex Hackathon!