There's certainly latent demand for a platform that localizes ad photography, although those customers are sensitive to weird artifacts in the generated images. Likely non-trivial r&d investment there.

The clearest immediate opportunity for GANs is generating content where artifacts might add value or are easily ignored (eg. art). The problem here is there's very little tech moat for these businesses given how easy it is to train a GAN. It'd come down to having a valuable, private dataset.

Lots of other potential commercial applications - we list some more on the demo.