ElevenLabs adopts Google's SynthID watermarking for all AI-generated audio
Tags AI · OSS · Policy
ElevenLabs has integrated Google DeepMind's SynthID invisible watermarking technology into its text-to-speech generations for free users, with plans to expand to all audio outputs in coming weeks. The watermarks can be detected using ElevenLabs' Audio Detector tool. The move addresses growing concern about AI-generated audio being used for impersonation and misinformation.
Technical significance
ElevenLabs' adoption of SynthID sets a precedent for proactive content authentication in generative AI audio. As voice cloning becomes trivially easy, invisible watermarking may become a regulatory requirement rather than a voluntary measure. The fact that Google is licensing SynthID broadly suggests it's positioning as a de facto standard for AI content provenance, similar to C2PA for images.