SolarRenderer renders realistic video from structural inputs in 4 steps via LoRA-adapted frozen video backbone
showlab · hf · 2026-10-21
NUS show lab and collaborators release SolarRenderer, a few-step generative reference rendering method: it adapts a frozen video backbone with lightweight LoRAs to render realistic videos from diverse structural inputs in just four steps, or in real-time streaming mode.
The key is frame-aligned training data. Project page, code, dataset, and weights are all open-sourced (GitHub: Enderfga/SolarRenderer).
More from Multimodal
- Developers pack Google offices for GenJam, making tiny videos with Nano Banana and Flow — DynamicWebPaige · 2026-10-21
- Controlling CapCut Web with Codex: a natural-language workflow for K-pop fan edits — SarahAnnabels · 2026-10-21
- 8 GitHub repos for making videos with Claude Opus 5.5, top collection at 7.2k stars — yihui_indie · 2026-10-21
- Gemini 4 Argon Builds a Pettable 3D Spider with 65,000 Fur Strands and Full Leg Physics — minchoi · 2026-10-21
- First Try: Gemini 4 Argon Recreates an Entire Video as Animated SVG in 20 Minutes — minchoi · 2026-10-21
- Gemini 4 Argon One-Shots Six Interactive Single-File 3D Demos with 3D-to-SVG Projection — minchoi · 2026-10-21