Music has changed in popularity over the years. From style to tempo and even popular instruments every aspect of music seems to change with every decade. While humans keep pushing towards a more technology driven life, we keep finding new ways to help us translate the rhythm inside so that others can enjoy it.
While some people might be on the fence saying that “Electronic music isn’t music” and “They don’t play REAL instruments” Humans are entering a time where it really only takes one person to create a song. By gone are the days of massive recording spaces for bands plus all their equipment to record a single or album.
High quality production can be achieved from a very basic setup and the right know how. But with advancing knowledge in the field of neural networks, artificial intelligence is slowly creeping its way into the music world.
Google has assembled a team called Magenta with the focus of creating art and music with machine learning. One project Magenta is working on is called Nsynth, short for Neural synth.
“Unlike a traditional synthesizer which generates audio from hand-designed components like oscillators and wavetables, NSynth uses deep neural networks to generate sounds at the level of individual samples. Learning directly from data, NSynth provides artists with intuitive control over timbre and dynamics and the ability to explore new sounds that would be difficult or impossible to produce with a hand-tuned synthesizer.”excerpt from Magenta’s NSynth website
While the Machine learning side try to get NSynth to produce a sound from learning in previous generations, the music side is looking for that musical quality; to see if a computer could actually produce music.
Nsynth uses data for the audio files to learn the fundamentals of that sound, taking that it can turn around and replay its own interpolation of the sound. With Nsynth you can also combine sounds/instruments to create some unconventional sounds. While the team at Magenta keeps feeding Nsynth audio to decode, AI generated audio is very rough.
Trying to simulate a person playing an instrument is difficult, the human nuances give the performance a special feeling knowing the piece was played with feeling and soul. You think the same could be said about singing too…
Hatsume Miku made her debut in the fall of 2007. Hailing from Japan and created by Crypton Future Media. Hatsume Miku, a 16 year old with Moe anthropomorphism. In reality she’s not real… but not fake because her voice exists but Hatsume Miku herself is not a real person but a vocaloid software.
Popular in anime and Japanese culture, Hatsume is actually voiced by Saki Fujita. The software created by Crypton Future Media takes those vocal samples and runs them through Yamaha’s Vocaloid. Vocaloid is a Concatenative synthesis program that focuses on the frequency domain. Initially, Vocaloid’s synthesis technology was called “Frequency-domain Singing Articulation Splicing and Shaping” but could be summed up by calling it singing articulation.
One popular song that has used a Vocaloid would be the 2017 single off Big Boi’s Boomiverse album “Kill Jill”. Stating the song you can hear Hatsume Miku singing while other instruments start to play.
Hatsume Miku is considered a Vst plug-in. A sell-able software that anyone can buy and use in anyone of their projects. But if you wanted artificial intelligence to help collaborate on your next album, it might be a little bit more tricky.
A lot of artificial intelligence is being created right now. The people that are working with it have a free open online platform you can play with their creation but it is all to feed the machine more and more data to teach it on how to perform the end goal. It isn’t necessarily hard to create your own neural network, but to create a music composer would be a challenge.
So while artificial intelligence may be taking its first steps into the music world, it won’t take long for the technology to produce a 100% A.I pop star.