Descubra a magia do machine learning nesta simulação casual que transforma mais de 44.000 amostras de áudio do Freesound em uma nuvem sonora 3D navegável. Os jogadores exploram os embeddings t-SNE do dataset, selecionam clipes marcantes e os reproduzem para criar instrumentos personalizados, como pianos a partir de notas synth ou marimbas com pops e estalos.
Mecânicas simples impulsionam uma criatividade infinita: mova-se pelo vasto mapa 3D, salve áudios selecionados e ative a reprodução para diversão rítmica. Esse experimento indie revela como o computador compreende dados sonoros, ideal para exploradores que preferem a descoberta à construção. Entusiastas de áudio e tinkerers casuais encontram um playground na sua escala imensa e variedade sonora.
Feel the magic of Machine Learning in this new experiment that combines over 44,000 sounds from Freesound, a data set created by Kyle McDonald and a touch of Cabbibo weirdness.
Run your hands through a cloud of sound, learn the shape of the computers mind, select your favorite audio delicacies and make a drum set from this decadent aural smorgasbord.
The mechanics may be simple ( navigate through a 3D map of sounds , select individual sounds to keep, and hit those sounds again to play them ), but because of absolutely massive size of the data set, the possibilities are endless!
Want to create your own piano out of synth notes? Go ahead!
How about a marimba of pops and crackles? Be my guest!
And if you are just the exploring type, the data is beautiful enough to begin with! There is no need to make an instrument, or play it, because in from of you lies a tool for discovering and understanding a beautiful algorithm, and even more beautiful data: Sound
This project would not be possible without the amazing community of FreeSound.org
As well as Kyle McDonald, who scraped audio from freesound.org & made t-SNE embeddings of the data!
Created during Unity Artist in Residency 2017
