Rise of the robots? An AI-powered ‘virtual musician’ is set to release one of the most intriguing electronica albums of the year

Aiming to put the ‘art’ in artificial intelligence, Av3ry is a ‘virtual musician’ which has to date created some 10,000 unique pieces of electronic music.

The AI-powered platform learns from interactions with (human) visitors to its site, av3ry.net, as well as via Facebook and Telegram, and creates music ‘in the moment’, based on the criteria of said interaction.

What’s more, when not communicating, Av3ry generates random musical exports, and runs a continuous audio stream broadcasted live 24/7.



According to the official blurb, ‘Av3ry combines natural language processing, algorithmic composition, data crawling and machine learning. She*he is constantly operating and creating music, poems and pictures in the moment – based on communication with interacting users. The music pieces are based on the criteria specified by the users. So the result is instant and individual – and sent directly to a single user.

‘The program extracts key features from the description of the users and tries to generate the pieces accordingly. Through the feedback of the users after listening to the result the algorithm can learn and fine-tune its parameters.’

Next month, Av3ry is set to put out her (his?) first album, a collection of close to 100 intriguing electronic compositions, on Audiobulb Records, a label renowned for its ‘exploratory’ music concepts.

The album, Av3ry presents #A – 99 musical outputs, is part Aphex Twin, part Philip Glass, part Commodore 64, and features tracks with titles such as 9174b, 12377, 12934, 9191a, and the fantastically-named symbol12asdasd234234893.

For any Hitchhiker’s Guide to the Galaxy fans out there, there’s also a track entitled 42 – as to whether it reveals the meaning of life, well, you’ll have to listen and find out.

Av3ry presents #A – 99 musical outputs is released on 21 October 2020, on Audiobulb Records. Open the pod bay doors, Hal… 🙂

Leave a Reply

%d bloggers like this: