Model Details
Wave2Vec2BERT and HuBERT are AI-synthesized audio detectors finetuned on the Wavefake dataset.
Developed by: TrustSafeAI
Model Sources
Project Page: https://huggingface.co/spaces/TrustSafeAI/AudioPerturber
Paper: https://arxiv.org/pdf/2503.17577
Uses
Users could use this detector to assist them in detecting audio generated by TTS models. Please note that the detectors are finetuned on AI-synthesized audio by Wave2Vec2BERT and HuBERT. As the models only supports non-commercial use, the intended users are not allowed to involve these detectors into commercial activities.
Ethical Considerations
we acknowledge that our detector is not perfect and will likely give incorrect predictions in some cases. In terms of ethical considerations, we suggest users use our tool to assist with identifying AI-synthesized auditory content at scale and with discretion. If the detection result is to be used as evidence, further validation steps are necessary as the detectors cannot always make correct predictions.