Kaleido model achieves state-of-the-art in neural rendering via unified 3D and video synthesis

Researchers have introduced Kaleido, a new family of generative models designed for neural rendering of objects and scenes. Kaleido treats 3D rendering as a sequence-to-sequence image synthesis task, enabling it to generate multiple views without explicit 3D representations. By leveraging large-scale video data for pre-training, Kaleido achieves state-of-the-art performance on view synthesis benchmarks, outperforming other generative methods in few-view scenarios and matching optimized methods in many-view settings. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel approach to neural rendering by unifying 3D and video modeling, potentially improving efficiency and quality in visual content generation.

RANK_REASON This is a research paper detailing a new generative model for neural rendering. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Kaleido
• Shikun Liu
• arXiv

• paper
• model release