Watermarking synthetic audio without retraining the model

Original: Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio

Source: arxiv.org ↗