---
title: "SpatialV2A: Visual-Guided High-fidelity Spatial Audio Generation"
authors:
  - "Yanan Wang"
  - "Linjie Ren"
  - "Zihao Li"
  - "Junyi Wang"
  - "Tian Gan"
doi: "10.24963/ijcai.2026/198"
url: "https://ijcai.org/proceedings/2026/198"
pdf: "https://ijcai.org/proceedings/2026/0198.pdf"
venue: "Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence"
year: 2026
pages: "1776-1784"
track: "Main Track"
keywords:
  - "Computer Vision: Image and video synthesis and generation"
  - "Computer Vision: Video analysis and understanding"
  - "Computer Vision: Multimodal learning"
---

# SpatialV2A: Visual-Guided High-fidelity Spatial Audio Generation

Yanan Wang, Linjie Ren, Zihao Li, Junyi Wang, Tian Gan

*Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence*, pages 1776-1784. DOI: [10.24963/ijcai.2026/198](https://doi.org/10.24963/ijcai.2026/198)

## Abstract

While video-to-audio generation has achieved remarkable progress in semantic and temporal alignment, most existing studies focus solely on these aspects, paying limited attention to the spatial perception and immersive quality of the synthesized audio. This limitation stems largely from current models' reliance on mono audio datasets, which lack the binaural spatial information needed to learn visual-to-spatial audio mappings. To address this gap, we introduce two key contributions: we construct BinauralVGGSound, the first large-scale video-binaural audio dataset designed to support spatially aware video-to-audio generation; and we propose an end-to-end spatial audio generation framework guided by visual cues that explicitly models spatial features. Our framework incorporates a visual-guided audio spatialization module that ensures the generated audio exhibits realistic spatial attributes and layered spatial depth while maintaining semantic and temporal alignment. Experiments show that our approach substantially outperforms state-of-the-art models in spatial fidelity and delivers a more immersive auditory experience, without sacrificing temporal or semantic consistency. All datasets, code, and model checkpoints are available at: https://github.com/renlinjie868-web/SpatialV2A.