Open Vision Agents by Stream. Build Vision Agents quickly with any model or video provider. Uses Stream's edge network for ultra-low latency.

End-to-end pipeline converting generative videos (Veo, Sora) to humanoid robot motions

Fast-FoundationStereo: Real-Time Zero-Shot Stereo Matching

The official repository for Nexels: Neurally-Textured Surfels for Real-Time Novel View Synthesis with Sparse Geometries

ACE-SLAM: Scene Coordinate Regression for Real-Time SLAM

Native and Compact Structured Latents for 3D Generation

Proof-of-concept surface reconstruction experiments to explore the design space for volumetric opaque solids.

Describe Anything, Anywhere, at Any Moment (DAAAM), a novel approach to real-time, large-scale, spatio-temporal memory

Code implementation for "Feedforward 3D Editing via Text-Steerable Image-to-3D"

TurboDiffusion: 100–200× Acceleration for Video Diffusion Models

Sharp Monocular View Synthesis in Less Than a Second

One-to-All Animation: Alignment-Free Character Animation and Image Pose Transfer

This project is a collection of Docker-based web user interfaces designed to easily run various state-of-the-art generative AI models locally. It simplifies the deployment of these AI tools by pack…

PersonaLive! : Expressive Portrait Image Animation for Live Streaming

[NeurIPS 25] TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels

ComfyUI nodes for SCAIL-Pose preprocessing

[NeurIPS 2025] Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance

Any4D: Unified Feed-Forward Metric 4D Reconstruction

Depth Anything 3

Official repo for paper "EMMA: Efficient Multimodal Understanding, Generation, and Editing with a Unified Architecture."

Training code for "Radiance Meshes for Volumetric Reconstruction".

This is an cut down version of our Vulkan based viewer to allow wider support.