NeoMap: Training-free Novel-View Synthesis from Single Images and Videos

2026-07-02

AI Summary

This work addresses the limitations of existing single-image or monocular video novel view synthesis methods, which often introduce artifacts and compromise global view consistency due to reliance on fine-tuning or rigid denoising guidance. The authors propose a training-free optimization framework that, for the first time, reveals the implicit capacity of pretrained video diffusion models to generate high-quality, viewpoint-consistent novel views. By iteratively refining initial noise through convergent manifold alternating projections within the natural video manifold, the method leverages only the pretrained model—without any additional training or camera-conditioned fine-tuning—to synthesize high-fidelity novel view videos. Extensive experiments demonstrate state-of-the-art performance on Tanks-and-Temples, LLFF, and DAVIS benchmarks, achieving superior generation fidelity and view consistency compared to current approaches.

Abstract

We study the challenging problem of novel view video synthesis from single images or monocular videos. Existing methods, which operate under the assumption that pre-trained video models lack native novel view synthesis capability and enforce view alignment via camera conditioning, task-specific fine-tuning, or stepwise hard denoising guidance, often suffer from artifacts and compromised global scene consistency. In this paper, we introduce NeoMap, a novel training-free framework designed to locate high-fidelity, view-consistent novel view solutions from general pre-trained video models. The key to our approach is the core insight that promising novel view solutions are inherently encoded within the natural video data manifold learned by pre-trained models, and the core challenge is simply to locate this optimal solution. We solve this via our core mechanism: convergent manifold alternating projection iterations that optimize the initial noise. Extensive experiments demonstrate that NeoMap significantly outperforms all existing methods across 3 standard novel view synthesis benchmarks, including the challenging Tanks-and-Temples, LLFF and DAVIS datasets, achieving state-of-the-art generation fidelity and top-tier view consistency.

Problem

• novel-view synthesis
• single image
• monocular video
• view consistency
• artifacts

Innovation

• training-free
• novel-view synthesis
• manifold projection
• pre-trained video models
• view consistency