Researchers have developed a method of making synthetic images more realistic using neural networks and the results when applied to GTA V are pretty damn impressive.
We present an approach to enhancing the realism of synthetic images. The images are enhanced by a convolutional network that leverages intermediate representations produced by conventional rendering pipelines. The network is trained via a novel adversarial objective, which provides strong supervision at multiple perceptual levels. We analyze scene layout distributions in commonly used datasets and find that they differ in important ways. We hypothesize that this is one of the causes of strong artifacts that can be observed in the results of many prior methods. To address this we propose a new strategy for sampling image patches during training. We also introduce multiple architectural improvements in the deep network modules used for photorealism enhancement. We confirm the benefits of our contributions in controlled experiments and report substantial gains in stability and realism in comparison to recent image-to-image translation methods and a variety of other baselines.
Unfortunately, you can't actually play the game in this photorealistic mode since it's only re-rendering the recorded footage. Maybe one day, but until then they need to apply the same methodology to Mortal Kombat because I've always wanted to watch video game footage and then throw up. Although finding the footage to train their model is going to be pretty difficult ever since LiveLeak shut down and took all their snuff films with it.
Keep going for the full demonstration video and check out side by side images on their project site.