Newest papers

SolarRenderer: Few-Step Generative Reference Rendering

Guian Fang, Shengju Qian, Dongyue Lu, Junchao Huang, Lutao Jiang, Xin Eric Wang, Yiren Song, Li Jiang, Mike Zheng Shou · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28630 · cited by 0 works (OpenAlex) · open-access PDF

German

Spielmotoren und Codierungsagenten können das Layout, die Bewegung und die Kamera einer Szene spezifizieren, aber eine solche grobe Struktur in ein realistisches Video zu verwandeln, erfordert einen Renderer, der der Struktur folgt und das Aussehen vervollständigt, das es offen lässt. Bestehende Renderer sind an die Darstellung eines einzelnen Motors gebunden, benötigen Dutzende von denoisierenden Schritten oder eine kostspielige Destillation ihres eigenen, und lernen von Bedingungen, die nicht mit ihren Zielvideos ausgerichtet sind. Wir präsentieren SolarRenderer, einen strukturbedingten Video-Renderer, der eine eingefrorene MiniMax-H3 33B-Rackbone mit zwei leichten LoRAs anpasst. Strukturvideos treten den nativen Referenzweg des Rack

Machine translation (Meta M2M100); the English original follows.

Abstract

Game engines and coding agents can specify the layout, motion, and camera of a scene, yet turning such coarse structure into realistic video requires a renderer that follows the structure and completes the appearance it leaves open. Existing renderers are tied to the representation of one engine, need tens of denoising steps or a costly distillation of their own, and learn from conditions that are not aligned with their target videos. We present SolarRenderer, a structure-conditioned video renderer that adapts a frozen MiniMax-H3 33B backbone with two lightweight LoRAs. Structural videos enter the native reference pathway of the backbone at the positions of the frames they describe, without a type identifier and described only in plain text, so that a single control LoRA follows arbitrary structural representations. Starting from a pretrained four-step LoRA, structure-guided distribution matching renders in four steps, 12.3 times faster than 50-step sampling, while preserving the learned control. A block-causal variant extends the same design to streaming and renders 768p video in real time. To train the model, we build a frame-aligned corpus of 136K clips from game engines, simulators, and YouTube videos. Experiments on an in-distribution test set and a generalization benchmark of 16 structural representations show that SolarRenderer achieves state-of-the-art results across both, and generalizes to 11 unseen representations, including pose skeletons, drawings, and point clouds.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Rendering (computer graphics), Artificial intelligence, Computer vision, Computer graphics (images), Coding (social sciences)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange