SolarWM goes fully open-source: 1.43M clips, 25TB pipeline for long-horizon video world models
Scobleizer · x · 2026-09-03
A team from CUHK-SZ, NUS, NVIDIA, MSRA and others has fully open-sourced SolarWM, an end-to-end pipeline for video world model research, which Scobleizer called "accelerant for robotics."
- SolarWM-Data: 1.43M video clips (25 TB) with camera annotations, quality scores, and the full curation/filtering stack.
- Long-horizon interaction: SolarWM-5B, trained only on 5-second clips, achieves real-time interactive generation with stable rollouts spanning minutes to hours.
- One framework, 4 backbones: unified training/inference for SolarWM-Wan-5B, Wan-14B, LTX-2.5, and Minimax-H3, including a causal student variant (4 steps).
Paper, code, checkpoints and datasets are all released as turnkey infrastructure for the community.
Related event: CUHK-SZ open-sources SolarWM long-horizon video world model(2 posts)→
More from Embodied
- French official meets Musk as Tesla FSD moves to on-road testing ahead of EU approval — elonmusk · 2026-09-03
- Tesla Cybercab certified at 165 Wh/mile, 31% more efficient than Model 3 — XFreeze · 2026-09-03
- First hands-on with vibecad: AI-generated molds for soft electrofluidic valves — IanPritchard · 2026-09-03
- London's first robotaxis launch on Uber: just 15 Wayve cars, safety driver onboard — nordicinst · 2026-09-03
- Langostino: open-source AI autopilot drone platform built on ROS2 — tom_doerr · 2026-09-03
- Petoi launches $99 open-source mini robot dog Quaddle for physical AI — Scobleizer · 2026-09-03