Descubre la magia del machine learning en esta simulación casual que transforma más de 44.000 muestras de audio de Freesound en una nube sonora 3D navegable. Los jugadores exploran las proyecciones t-SNE del dataset, seleccionan clips destacados y los reproducen para crear instrumentos personalizados, como pianos a partir de notas synth o marimbas con pops y crujidos.
Mecánicas sencillas impulsan una creatividad infinita: recorre el vasto mapa 3D, guarda audios seleccionados y activa la reproducción para ritmos divertidos. Este experimento indie pone al descubierto el dominio de la máquina sobre los datos sonoros, ideal para exploradores que prefieren el descubrimiento a la construcción. Audioaficionados y curiosos casuales hallan un patio de juegos en su inmensa escala y variedad sónica.
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
