Spatial sound simulation aims to produce sound with the characteristic of virtual scenes. Visual information attached with spatial sound enhances users' experience in virtual reality. To make users feel immersive in environments, we compute dynamically the sound filters according to the structure of scene instead of playing pre-recorded audio files. Our research uses an image source method to generate sound propagation paths in scenes. To improve the performance of producing image sources, we pre-computed the visibility of scene, and exploited Optix ray tracing engine to parallelize path validation. Furthermore, our system can simultaneously support multiple sound sources.