I don't think it will be feasible for a while if we consider the traditional means for developing a game, decades at least, probably more though. It takes massive teams of hundreds of people years to create games like RDR2 or Starfield, and those games still aren't quite photorealistic. Making a photorealistic rock or something isn't particularly difficult, but filling a massive open world with thousands of rocks, plants, trees, buildings, etc. takes a lot of time and manpower that simply makes it infeasible to develop at the moment.

However, I suspect there will eventually be a point where AI takes over the video game development process. AI will almost definitely be able to develop games so realistic that we won't be able to tell the difference, and it will probably be able to do it much quicker than humans. Theoretically if said AI has access to quantum computers thousands or millions of times faster than standard PCs we use today, then it could develop such a game in the blink of an eye, though it would require an unfathomable amount of computing power.

That being said, I think our GPU and general computing capabilities aren't quite there yet to support such a game, even if humans or AI were able to create it.

And certainly our AI is a ways away from being able to do that, though once we reach the theorized inflection point which Nvidia recently claimed we're close to, then all bets are off. Computing capabilities could advance so quickly that in 5 years we could have GPUs thousands of times more powerful than a 4090 today.

So TLDR; I think if Jensen Huang is right, about how fast AI is developing then it won't be very long for games to become indistinguishable from reality. If he's wrong then it will be another 30 or 40 years at least.