A new project aiming to enhance the realism of ‘synthetic images’ has gone viral after boosting Grand Theft Five’s Los Santos to photorealistic levels.
The age-old argument that ‘graphics don’t make a game’ has always been true. Games can still deliver fantastic experiences regardless of how they might look. Though every now and then, new games come along that push the visual boundaries of what’s possible in the medium.
When GTA 5 launched in 2013, it did just that, marking a major graphical leap for Rockstar Games. It was an exceedingly ambitious project for the given hardware and thanks to new ports, one that has stood the test of time. While Los Santos still looks stunning to this day, you’ve never seen it quite like this thanks to a new photorealistic enhancement.
A brand new development aimed at improving the quality of images has been applied to footage of GTA 5. As a result, we now have a shocking look at a GTA title running with “photorealistic accuracy” to the real world.
Through Intel Labs, an intern group put together a ‘convolutional network’ that drastically improves the quality of any given image. One frame at a time, this program is assessing incoming data from GTA and spitting back out an enhanced version.
The end result leads to remarkably authentic visuals that mirror the outside world as we know it. Driving through the upgraded streets of Los Santos looks as if you were driving through a typical American suburb.
Side-by-side comparisons showcase the true power of this new tool, as GTA 5 is modified to match the real world.
- Read More: NoPixel GTA RP community mourns roleplayer
Lighting is stunning, textures are true to life, and the appearance is better than ever before. Given how intense processing would be for tech of this caliber, however, don’t expect to see it in standard gaming anytime soon.
This project is limited to adapting footage only, not moment-to-moment gameplay. Who knows how far off we are from painstakingly realistic gameplay quite like this.
- Read More: When is GTA 6 coming out?
If you’re interested in the science behind it all, you can read through the full 16-page paper right here, courtesy of the Intel Labs group.