Google Magenta introduces its latest project that features a customizable virtual music room experience powered by its machine learning models based on the Tensor Flow.js, the Lo-Fi Player.
The Lo-Fi Player gives users a virtual music room, allowing them to relax and enjoy the music it creates. It features Lo-Fi music, allowing users to explore, customize, and produce their own beats. It works simply by clicking the elements found in the room to customize some of its visuals as well as the musical instruments and sounds based on the user’s preference.  Other things that users may change are the view on the window, the color of the wall, the pet to either a cat or a dog, the board which texts can be edited, and other clickable elements.
The Lo-Fi Player incorporates various Magenta music machine learning models including Melody RNN on the room’s radio which generates new melody, MusicVAE on the TV for melody interpolation and mixing two melodies together.
Moreover, Lo-Fi Player allows users to share their musical room with others through an interactive YouTube stream. The YouTube stream works by typing commands into the Live Chat instead of clicking the elements in the room. The commands allow the user to change the color of the room, change the melody, switch the instruments, and more. The commands will be randomly selected by the system and once chosen, the comment will be highlighted. The changes take effect after the beat loops. This allows users to collaborate with other’s musical preferences.
There is also another virtual musical experience that developers may get to explore which is the Play, Magenta!. It allows editing sounds and canvas with the editor live in their browser through source codes.
The Lo-Fi Player was built by Vibert Thio, an intern on the Google Brain team. You may try the Lo-Fi Player here.
Source: Google Magenta
