Google successfully integrated “watermarks” into artificial intelligence music creation, making them difficult to detect by human ears

IT House News on November 16, DeepMind, an artificial intelligence company owned by Google, announced that the AI ​​audio generation model Lyria it developed will use SynthID technology to add "watermarks" to the generated audio so that people can identify them afterwards. Made by AI.

DeepMind pointed out in a blog post that this watermark cannot be detected by human hearing and will not affect the listening experience. Watermarks can still be detected even if the audio is compressed, sped up, slowed down or extra noise is added

DeepMind stated that SynthID’s audio watermarking works by converting the audio waveform into a two-dimensional image, showing the changes in the sound spectrum over time. DeepMind claims this approach is unlike any other watermarking technology currently in existence

Watermarking tools such as SynthID are seen as an important safeguard against harm caused by generative AI. This is a promising field, but current technology is far from being a master key against counterfeiting.

The above is the detailed content of Google successfully integrated “watermarks” into artificial intelligence music creation, making them difficult to detect by human ears. For more information, please follow other related articles on the PHP Chinese website!