Google is embedding inaudible watermarks right into its AI generated music::Audio created using Google DeepMind’s AI Lyria model will be watermarked with SynthID to let people identify its AI-generated origins after the fact.
Google is embedding inaudible watermarks right into its AI generated music::Audio created using Google DeepMind’s AI Lyria model will be watermarked with SynthID to let people identify its AI-generated origins after the fact.
Like one can use POV-Ray for pictures, one can use a script generating ABC sequences and choosing keys and instruments according to some math, which are then converted to MIDI, for music. It’s not all machine learning (which is more or less just extrapolation).
Then, with ML you can be smarter than just finding a generator and feeding it keywords. You can (with converters and stuff) use ML for sequences and keys and instruments known to “feel good” with your own brain. Say, playing with different optimization criteria for the last N elements (notes) of a sequence (using how they relate to each other etc). EDIT: … and then use the result to generate something.
I guarantee that the result won’t be anywhere near perfect and beautiful, though.