tech
Google is embedding inaudible watermarks right into its AI generated music
Posts from this topic will be added to your daily email digest and your homepage feed.
TL;DR
- Google DeepMind's AI Lyria model and YouTube's audio features will use SynthID to watermark AI-generated audio.
- The watermark is designed to be inaudible and not compromise the listening experience.
- SynthID aims to be detectable even after audio compression, speed changes, or added noise.
- Watermarking is considered an important safeguard against generative AI harms, with government standards being developed.
- The audio implementation of SynthID visualizes the audio wave's frequency spectrum over time.
- Google previously released SynthID for images generated by its Imagen model, which is resistant to some image manipulations.