OfSpectrum

OfSpectrum offers proprietary technologies focused on audio watermarking, anti-AI encryption, and deepfake detection for audio content. Its solutions address the need to verify the authenticity of audio, prevent unauthorized use or cloning of voices, and identify synthetic audio generated by AI. The platform is designed for those seeking to protect audio assets from piracy, unauthorized AI training, and deepfake threats.

The audio watermarking technology developed by OfSpectrum embeds inaudible multimedia data—including text, audio, images, or links to online databases—directly into audio files. This embedded data enables verification of the audio's origin and authenticity, supporting piracy prevention and content verification efforts. The watermarking is robust against common editing and compression, and it also allows for private channel distribution of multimedia data.

OfSpectrum's anti-AI encryption applies inaudible modifications to audio, making it difficult for AI models to clone or replicate voices. The encryption is designed to disrupt both the data crawling process used for AI training and instant voice cloning attempts during inference. Each encryption instance is unique, which increases its resistance to bypass or removal. The technology supports high-quality audio up to a 48kHz sample rate without introducing distortion.

For detecting synthetic audio, OfSpectrum provides deepfake detection capabilities that distinguish between AI-generated and real human voices in real time. The underlying model has received multiple awards in international audio deepfake detection competitions, highlighting its reliability and performance.

The platform appears to offer access to its technologies through an API, as indicated by references to open access API documentation.