Ph.D. candidate @HKUST | videogen, generative rendering, robotics
Joined October 2022
- Tweets22
- Following67
- Followers30
- Likes42
love this demo
Reactor is now running in @Blender ! VAST's EdenLive is a perfect fit for running realtime AI video directly within the viewport. It's so easy to try different styles using the same consistent environment geometry. Send me a DM if you want to install the plugin yourself!
1
55
I think we’re already starting to see the early form of the next-gen rendering pipeline—It’s definitely worth a try.
EdenLive, a real-time generative renderer, is now open-sourced by @VASTAIResearch.
It transforms 3D blockouts, game engine footage, 2D cartoons, and real-world video into detailed photorealistic or stylized video.
24+ FPS at 720p. A 5B model on a single RTX 5090.
1
71
DeepSeek always makes everything open source🫡
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
66
A static sky may need just one denoising step. A moving subject may need four. Yet most video diffusion models give every patch the same budget.
🚀 Introducing ForcingBooster: adaptive patch-wise denoising that lets easy regions finish early while spending more computation where refinement is still needed.
1/4
ForcingBooster cuts denoising steps across video patches by ~50% and delivers a 1.30× end-to-end speedup, with visual quality comparable to the four-step baseline.
1
63
ForcingBooster is officially released!
Arxiv: arxiv.org/abs/2610.12625
Website: igl-hkust.github.io/ForcingB…
Code: github.com/IGL-HKUST/Forcing…
Models: huggingface.co/EXCAI/Forcing…
1
60
Excited to share that our work, GO-Renderer, has been accepted to #ECCV2026! 🎉
Looking forward to meeting and discussing with everyone at ECCV in Sweden! 🇸🇪
📄 Paper: arxiv.org/abs/2603.23246
💻 Code: github.com/IGL-HKUST/GO-Rend…
#ECCV2026 #3DVision #VideoDiffusion #GenerativeAI
9
323
Zekai Gu retweeted
Sharing our recent work, GO-Renderer, which leverages generative models to perform high-quality controllable object rendering from sparse images without exact geometry and appearance modeling.
Homepage: igl-hkust.github.io/GO-Rende…
Paper: arxiv.org/abs/2603.23246
Zekai Gu retweeted
Diffusion as Shader (DaS) is accepted by SIGGRAPH'25, which enables multiple video creation tasks using 3D control signals. DaS generates all videos below.
The codes have already been released! Welcome to try DaS! github.com/IGL-HKUST/Diffusi…
I'm excited to share our new work Diffusion as Shader (DaS), a versatile controllable video generation method for various tasks: object manipulation, camera control, mesh-to-video, and motion transfer.
Project page: igl-hkust.github.io/das/
Github: github.com/IGL-HKUST/Diffusi…
Zekai Gu retweeted
We discovered that imposing a spatio-temporal weight space via LoRAs on DIT-based video models unlocks powerful customization!
It captures dynamic concepts with precision and even enables composition of multiple videos together!🎥✨
Zekai Gu retweeted
I'm excited to share our new work Diffusion as Shader (DaS), a versatile controllable video generation method for various tasks: object manipulation, camera control, mesh-to-video, and motion transfer.
Project page: igl-hkust.github.io/das/
Github: github.com/IGL-HKUST/Diffusi…