It’s no secret that AI-based technology has been quickly growing in the past few years. From self-driving cars, to Nvidia’s DLSS, AI can be used on a wide variety of different technologies for various different reasons. One AI research group has attempted to create a photorealistic image enhancer, which can be used to make some games look almost unrecognizable.

The research group known as Intel ISL has been working on a photorealistic enhancement tool that uses machine learning to improve a game’s quality. Essentially, it takes an image of the game and compares it to a real-life one, then applies some adjustments based on the real-life image to the original. In Grand Theft Auto 5, the results are pretty incredible:

Now there are obviously some kinks to be ironed out with this, as it’s not exactly perfect just yet (most of the real-world images the tool uses to compare are tinted low-quality dashcam footage, hence the 90’s-style color grading).

However, you can immediately see the difference in certain places, like how the road’s asphalt looks a lot smoother, or how cars become more reflective, resulting in a more photorealistic image. You can also have a look at some more comparison images on their official web page.

But what’s more interesting is that the technique the tool uses can apparently produce enhanced images at “interactive rates,” which essentially means it can work in real time. Theoretically that means you could apply it to the game yourself, though Intel ISL have yet to release the official tools used in the video above.

Maybe this could help games get more of a photorealistic look in the future, or maybe it will just be saved for fun mods to improve a game’s textures. Either way, it certainly is cool to watch. But what do you think? How do you feel about the enhancements above? Does GTA 5 look better? Or is there still a lot of work to be done? And what could this technology be used in the future for? Let us know your thoughts!

Vote - Click on the bar or text you want to cast your vote on
Vote - Click on the bar or text you want to cast your vote on
Vote - Click on the bar or text you want to cast your vote on