Experiments

A collection of rough ideas and tests

StyleGAN2 encoder – Finding doppelganger in an AI generative model

source code: https://github.com/rolux/stylegan2encoder/

M4L.RhythmVAE – Train and use AI rhythm generators on Ableton Live

Variational Autoencoder(VAE)-based rhythm generation device for Ableton Live/Max for Live. This plug-in software allows artists/musicians to train and use the rhythm generation model within the music production software(DAW).

AI-Generated “Music” / 楽譜からうまれるメロディ

Made in collaboration with Japan Graphic Designers’ Association (JAGDA).http://qosmo.jp/projects/jagda-exhibition-2/

M4L.RhythmVAE: VAE Rhythm Generator for Ableton Live

Variational Autoencoder(VAE)-based rhythm generation device for Ableton Live/Max for Live. This plug-in software allows artists/musicians to train and use the rhythm generation model on the music production software(DAW). Users of the device can drag and drop MIDI files containing rhythm track and train the model. Once trained, they can automate the transition of the latent…

“Music” generation with GANs

"Music" generation with GANs pic.twitter.com/MJKXPpGlXX — Nao Tokui (@naotokui_en) March 15, 2019 Trained a GAN model on CVC-MUSCIMA hand-written music score dataset.    Alicia Fornés, Anjan Dutta, Albert Gordo, Josep Lladós. CVC-MUSCIMA: A Ground-truth of Handwritten Music Score Images for Writer Identification and Staff Removal. International Journal on Document Analysis and Recognition, Volume 15, Issue 3,…

AI DJ Project at Google I/O 2019 preshow

  We were invited to perform our AI DJ Project at the opening of Google I/O 2019! (Video: @gnoot)

Generating drum patterns with Variational Autoencoder [work in progress]

 

Neural Beatbox (alpha)

Try it by yourself: https://codepen.io/naotokui/pen/NBzJMW Make some noise and record your voice! Deep Learning models will analyze and build a drum kit, then start generating drum patterns infinitely with your sound. Convolutional Neural Network is used to analyze and classify audio segments based on spectrograms and Recurrent Neural Network(LSTM) for generating drum sequences. Shout-out to…

Imaginary Landscape at ICC OpenSpace 2018

A quick video documentation of my new installation “Imaginary Landscape” exhibited at NTT InterCommunication Center(Tokyo).   Images of landscapes that together look like one continuous scenery are projected onto screens on three walls of the exhibition space. The work is realized by analyzing photographs found in Google Street View, selecting three pictures each that look similar…

Imaginary Soundscape – website

Now you can upload images and listen to the AI-generated soundscape on our website!  Robin Jungers, who joined Qosmo recently, worked on this project. http://imaginarysoundscape2.qosmo.jp/