Objectives: To train participants to localize sound using virtual reality (VR) technology, appropriate auditory stimuli that contain accurate spatial cues are essential. The generic head-related transfer function (HRTF) that grounds the programmed spatial audio in VR does not reflect individual variation in monaural spatial cues, which is critical for auditory spatial perception in patients with single-sided deafness (SSD). As binaural difference cues are unavailable, auditory spatial perception is a typical problem in the SSD population that warrants intervention. This study assessed the applicability of binaurally recorded auditory stimuli in VR-based training for sound localization in SSD patients. Methods: Sixteen subjects with SSD and thirty-eight normal-hearing controls (NHs) underwent VR-based training for sound localization and were assessed three weeks after completing training. The VR program incorporated prerecorded auditory stimuli created individually in the SSD group and over an anthropometric model in the NH group. Results: Sound localization performance revealed significant improvement in both groups after training, with retained benefits lasting for an additional three weeks. Subjective improvements in spatial hearing were confirmed in the SSD group. Conclusion: By examining individuals with SSD and NH, VR-based training for sound localization that used binaurally recorded stimuli, measured individually was presented as effective and beneficial. Furthermore, VR-based training does not require sophisticated instruments or setups. These results suggest that this technique represents a new therapeutic treatment for impaired sound localization.