In this chapter, we first gave an overview of different neural-style transfer methods developed since 2015. Then we showed how to train a second-generation style transfer model that's fast enough to run on mobile devices in a few seconds. After that, we covered how to use the model in both an iOS app and an Android app, built from scratch with a minimalist approach with fewer than 100 total lines of code. Finally, we talked about how to use a TensorFlow Magenta multi-style neural transfer model, which includes 26 amazing art styles in a single small model, in both iOS and Android apps.

In the next chapter, we'll explore another task deemed as intelligent when demonstrated by us humans or our best friends: to be able to recognize voice commands. Who wouldn't want our dogs to understand commands such as "sit," "come," "no," or...