April 2021/100% neural net synthesized soul/funk. 12 hours of it, free to download and remix.
Dadabots and Keyon Christ have partnered to bring to life a new incarnation of soul music as imagined through creative AI and human ingenuity.
2021
Keyon Christ describes this work, “Out of all the major songs I’ve worked on, Black Skin Machine is the coolest. I believe it’s the first song in the world that extracts this much soul from AI.“
NO SOUL from a bird's-eye view for hunting and gathering. Find the perfect gems in the Disproportionately Oversized Music Explorer. Select and press 'e' to export.
Click D.O.M. E. Image to Start Exploring
NO SOUL sample pack 2.38 GB [download] contains 12 hours of full tracks, synthesized with OpenAI Jukebox, mastered using Landr, and stemmed into bass/vocals/drums/other using Spleeter. It’s free to download, redistribute, and remix.
NO SOUL 24/7 livestream. It dreams basslines, choirs, changes, buildups, glitches, and surprises. At each moment, it’s evolving and unpredictable. It’s synthesized, yet it’s got us feeling good, headbanging, and at times, full-on stankfacing.
Remixing funk records is the origin of hip hop. Yet in 1992, artists started getting sued for sampling. Neural synthesis gives back to music producers the freedom to remix. We can now synthesize music that quotes a particular era, without sampling any previously published recording.
We set up a distributed system of 36 NVIDIA V100 GPUs to generate thousands of tracks in parallel, exploring various temperature and hop_size hyperparameters conditioned on funk combined with various other style vectors. The model has listened to thousands of artists, and its style draws inspiration across hundreds of subgenres.
CJ Carr
For hours every morning in his basement, CJ practices blast beats, training for the day he’ll need to represent humanity in a drum-off against the machines. Despite his practice, CJ realized the easiest way to collaborate with his favorite musicians was to create algorithms for them to use. He has competed in 65+ hackathons, which is way too many all-nighters drinking tea and coding. CJ’s day job is CTO at a hip hop record label building hybrid audio/AI/EEG inventions.
Zack Zukowski
"It's like baking cookies" — Zack, waiting for the first epoch to finish training.
Zack is a music technologist and multi-instrumentalist. In high school, Zack toured with a punk band. He started creating music bots after meeting CJ at Berklee College, where they had worked on music-education games. Zack approached his machine learning studies from an audio perspective, presenting his research on neural music synthesis at NIPS 2017, MUME 2018, and MILC 2019. Now, he manages the machine learning team at Pex, an audio/video search engine. Zack falls asleep at night imagining an absurd future of bots remixing bots, generating music only for machines to listen to in vast abandoned sections of the Internet.
Twitter | Instagram | YouTube
Keyon Christ, formerly known as Mitus, is a Los Angeles-via-Atlanta futurist meditating in the realms of technology and music production. At age 19, Christ landed placements with Rihanna on the industry-rattling ANTI album (Goodnight Gotham), as well as Kanye’s grandiose The Life of Pablo (FML feat. The Weeknd), both of which received Grammy nominations as well as platinum certifications by the RIAA. With aims of furthering his own creative autonomy, Keyon dropped his debut They Don’t Want Us EP in 2019. Besides furthering his own artistic endeavors, Keyon is continuing to secretly shape the next generation of music technology.
Twitter | Instagram | Website
DADABOTS – CJ Carr and Zack Zukowski - are two hacker-musicians taking Deep Learning to generative music. Their work – the products of large datasets of recorded audio, processed to produce new music – creates a new meaning for ‘sampling’. Their 24/7 livestreams - some running non-stop for 2 years now - constantly synthesize music with neural nets, from datasets of Math Rock, Free Jazz, Death Metal, Funk, and even the Beatles, challenging the listener to reinterpret their understanding of human-created music. Their series of Impossible Covers creates style-transferred songs from pop canon - performed by other artists, such as Frank Sinatra Sings Britney Spears and Nirvana Cover Gorillaz.
This talk will feature an overview of their technical work and toolsets, a dive into their artistic exploration, and brand new work with hip-hop producer Keyon Christ.
Dataset processing isn't new to musicians; entire musical genres are based solely on manipulating samples. Musical production techniques have become so sophisticated that listeners only perceive musical textures nowadays. But behind the scenes, the methods of making those textures are changing dramatically. This panel will feature creators and users of electronic instruments that rely on deep learning for their performance — whether to create entirely new sounds and music from existing recordings, or to give the music playing a human form.