Oriel — AI & technology worth noticing, every day.

SolarRenderer: Few-Step Generative Reference Rendering

Paper · https://arxiv.org/abs/2610.28630

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.

First seen by Oriel on .

AI summary

SolarRenderer adapts a frozen MiniMax-H3 33B backbone into a structure-conditioned video renderer that generates realistic video in four steps and generalizes across diverse scene representations.

  • It uses two lightweight LoRAs, placing structural videos in the backbone’s native reference pathway and describing them only with plain text.
  • Structure-guided distribution matching preserves control while rendering in four steps, 12.3 times faster than 50-step sampling.
  • A block-causal variant supports streaming and real-time 768p video generation.
  • The training corpus contains 136K frame-aligned clips from game engines, simulators, and YouTube videos.
  • On a benchmark covering 16 structural representations, it reports state-of-the-art results and generalizes to 11 unseen types, including pose skeletons, drawings, and point clouds.

#video generation #generative rendering #lora #game engines

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 53, latest observation ; +13 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 2, latest observation ; +1 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2610.28630https://arxiv.org/abs/2610.28630 Read discussion