Aiva Technologies, the company behind the world’s first non-human AI music composition, just released a brand new AI-generated rock music track. The work builds on the company’s previous compositions that range from cinematic film scores to Chinese music.
Using NVIDIA TITAN Xp GPUs, GeForce GTX 1080 Ti GPUs, the company re-trained their deep neural network on several hundred rock tracks. The company already has a database of 30,000 scores but it is largely made up of classical, and symphonic music encoded in the MIDI format.
“Melody is key to a really great Rock track, and at first, we didn’t expect anything super interesting to happen on that front. We did expect to have some cool harmonies that would give a nice starting point to a great Rock song, but nothing that would really stand on its own,” the Aiva team wrote in a blog post.
Once they re-trained their neural network, the company relied on NVIDIA Tesla V100 GPUs on the Amazon Web Services Cloud to compose the track. The result is a rock track called On The Edge.
“Sometimes, it can be hard for our neural networks to capture the mechanics of great melodies in styles that are less rich than classical music. But our AI music composer AIVA proved us wrong, and we’re happy to say the result was above our expectations!
The team says the raw track generated by the algorithm had some weaknesses, such as a melody in the bridge but they were able to improve the song by removing the melody from that part of the song.
“We’re very excited about the results achieved with On the Edge, because we want to allow composers to collaborate with AIVA in a wide variety of genres,” the team said.
Developers looking to use the application to compose music can sign up for the company’s beta release.
Read more>
Related resources
- GTC session: AI Will Never Write a Hit Song. Or Will It??
- GTC session: Generative AI Theater: AI Decoded - Generative AI Spotlight Art With RTX PCs and Workstations
- GTC session: AI-Driven Generative Music for Singing Voice Conversion and DAW-Less Operation with an 18,000-Song Karaoke Catalog
- NGC Containers: MATLAB
- NGC Containers: Morpheus Triton Server Models
- NGC Containers: vst