In this chapter, we've looked at Tensorflow.js and Magenta.js, the JavaScript implementations of TensorFlow and Magenta. We've learned that TensorFlow.js is GPU accelerated using WebGL and that Magenta.js has a limited set of models available that can only be used for generation, not training. We've converted a Python-trained model from the previous chapter to a format that TensorFlow.js can load. We've also introduced Tone.js and the Web Audio API, which is used by Magenta.js to synthesize sound in the browser.
Then, we've created three music generation web applications. The first application used GANSynth to sample short audio notes. By doing so, we've learned how to import the required scripts, either using a big ES5 bundle or a smaller, split up, ES6 bundle. The second application used MusicVAE to sample a trio of instruments, with the...
