Here the World in Stereo: Learning Dynamic Spatial Correspondence for Immersive Joint Video-Audio Generation
StereoBind ties moving visual sources to stereo sound so the audio position follows the object on screen.
The paper targets joint video-audio generation for AR/VR and gaming, where synchronized sound is not enough if stereo placement drifts. Its framework uses motion tracks to bind source-aware visual motion, absolute source position, and relative movement into stereo audio generation. The authors also introduce StereoWorld-29K for supervision and StereoWorldBench to evaluate spatial consistency. They report better stereo spatial alignment than existing models while maintaining overall audiovisual quality. HF Daily Papers' note
The paper targets joint video-audio generation for AR/VR and gaming, where synchronized sound is not enough if stereo placement drifts. Its framework uses motion tracks to bind source-aware visual motion, absolute source position, and relative movement into stereo audio generation. The authors also introduce StereoWorld-29K for supervision and StereoWorldBench to evaluate spatial consistency. They report better stereo spatial alignment than existing models while maintaining overall audiovisual quality. HF Daily Papers' note
score 4