SynSFX: Multi-Model Sound Effects Synthesis Dataset for Deepfake Detection and Evaluation
2026-07-06 • Sound
SoundArtificial Intelligence
AI summaryⓘ
The authors study the challenge of detecting fake sound effects generated by computers. They note that current detectors don’t work well on synthetic sound effects and that existing datasets are small or not well-suited for this problem. To improve this, the authors created SynSFX, a large new dataset with over 43,000 clips, including both real and synthetic sounds from popular text-to-audio models. This dataset helps researchers better understand and detect fake sound effects.
audio deepfake detectionsynthetic sound effectstext-to-audio modelsdatasetEnvSDDmachine learningsound generationaudio forensics
Authors
Linxi Li, Yuncong Yu, Qianwei Guo, Liwei Jin, Yechen Wang, Carsten Maple
Abstract
While audio deepfake detection has advanced significantly, representative detectors show limited generalization to synthetic sound effects. Existing environmental audio datasets such as EnvSDD provide important initial resources, but remain limited in scale and generation provenance for studying isolated sound-effect deepfakes. To support this direction, we present SynSFX, a large-scale corpus of 43374 clips (26452 synthetic, 16922 real) spanning 7 popular text-to-audio models.