GANs are indeed interesting networks; besides the applications you've seen, GANs have been explored in many other exciting applications. In the following, we list a few:
- Music generation: MIDINet, a convolutional GAN, has been demonstrated to generate melodies. You can refer to the paper here: https://arxiv.org/pdf/1703.10847.pdf.
- Medical anomaly detection: AnoGAN is a DCGAN shown by Thomas Schlegl et al. to learn a manifold of normal anatomical variability. They were able to train the network to label anomalies on optical coherence tomography images of the retina. If the work interests you, you can see the related paper on arXiv at https://arxiv.org/pdf/1703.05921.pdf.
- Vector arithmetic on faces using GANs: In the joint ...