arXiv:2605.31535v1 Announce Type: cross 
Abstract: Self-supervised novel view synthesis (NVS) remains challenging to scale, despite the abundance of video data, largely due to the brittleness of training on realistic videos and the hard-to-predict scaling behavior of multi-network system designs. We introduce RayDer, a unified, feed-forward transformer that consolidates camera estimation, scene reconstruction, and rendering into a single backbone, turning self-supervised NVS into a well-posed single-model scaling problem. A minimal dynamic state, treated as a nuisance factor, absorbs time-varying content and enables stable training on unconstrained real-world video. Importantly, RayDer keeps static-scene NVS as its target task: dynamic content is leveraged purely as scalable supervision, not reconstructed as in dynamic-scene (4D) NVS. Across multiple model sizes and orders of magnitude in data, RayDer exhibits clean power-law scaling with data and compute, and outperforms static-scene data mixtures. On a large number of benchmarks, RayDer achieves strong zero-shot open-set performance competitive with state-of-the-art supervised approaches. Project Page: https://compvis.github.io/rayder

يمثل تقديم RayDer، وهو محول موحد ذو تغذية أمامية، تقدمًا كبيرًا في تركيب وجهات نظر جديدة ذاتية الإشراف (NVS) من مقاطع الفيديو الواقعية. يجمع هذا النموذج بين تقدير الكاميرا وإعادة بناء المشهد والتصيير في إطار واحد، مما يعالج التحديات المتعلقة بتوسيع نطاق NVS على الرغم من توفر بيانات الفيديو الواسعة.

La introducción de RayDer, un transformador unificado de avance directo, marca un avance significativo en la síntesis de nuevas vistas auto-supervisada (NVS) a partir de videos del mundo real. Este modelo consolida la estimación de la cámara, la reconstrucción de escenas y el renderizado en un solo marco, abordando los desafíos de escalar la NVS a pesar de la disponibilidad de amplios datos de video.

L'introduction de RayDer, un transformateur feed-forward unifié, marque une avancée significative dans la synthèse de nouvelles vues auto-supervisée (NVS) à partir de vidéos du monde réel. Ce modèle regroupe l'estimation de la caméra, la reconstruction de scène et le rendu en un seul cadre, répondant aux défis de mise à l'échelle de la NVS malgré la disponibilité de vastes données vidéo.

The introduction of RayDer, a unified feed-forward transformer, marks a significant advancement in self-supervised novel view synthesis (NVS) from real-world video. This model consolidates camera estimation, scene reconstruction, and rendering into a single framework, addressing the challenges of scaling NVS despite the availability of extensive video data.