Frame Generation is an AI tool that insert "fake frames" between "real frames" in order to boost FPS in games.

Those frames are generated in order so you can get more smooth gameplay without having to upgrade your machine. But is that the right thing to do?

Frame Gen is usually bashed around the gamers, since it's AI generated, and not "real frames".

What do you think? Have you ever used Frame Generation? What is your opinion on it?

I haven't used it in a ton of games yet, but in the games I have, it's worked out pretty well so far. As long as it doesn't add too much input lag, why wouldn't I want to use it, since it generally seems to look OK? Besides, as far as I know, it's really just a particularly good interpolation method. We could probably have had frame generation way before AI, only it would have looked worse.

The part I don't love about frame generation is that it enables developers to cheap out on optimization. I don't know much of that actually happens, but on some time scale, it undeniably does or will lessen the pressure to optimize games further. But since frame generation is here, for better or worse, I might as well utilize it. I guess if I hit the refresh rate of my monitor, I could skip using frame generation, but that doesn't really happen a lot, at least not without significant sacrifices in other areas.

I don’t see what meaningful labor is lost from allowing frames to be more easily generated (or rather, I don’t see how it would be unethical). Seems like an appropriate use of technology. Can’t just blindly be opposed to any use of the “new and scary” innovation. There has to be some shades of grey in there.

Though I am all ears to hearing how such could be problematic. It’s not something I’ve looked into extensively so I’ll happily be informed by somebody looking to explain.

I was initially sceptical of it due to the input latency impact but now that I've tried it I think the tech is so incredible that it feels like magic. Said latency impact is actually small and for that slight sacrifice you get a massive boost to visual smoothness so it's perfect for playing a demanding game on a display that is 240hz or higher.

I tried it properly for the first time in Forza Horizon 6 and thought it worked super well there though was still unsure how well it'd work with something like a fast paced FPS but 4x frame gen let me play Doom: The Dark Ages at 240fps max settings aside from path tracing with basically no issue aside from the input latency which was hard to notice so I am now a massive believer in it.

It'll be how super high refresh rate displays like 1000hz ones will be able to show that many frames for games that are even remotely demanding. DLSS upscaling is also incredible for letting you get close to 4K image quality from native 1080p but frame gen might be even more impressive and magic like to me. It'll be how the next gen consoles will be able to consistently offer 120fps modes and in the coming years it'll be possible to use it to go from 100fps to 1000fps with little drawback. It is kinda a rich gets richer feature though since the higher the base frame rate is the better, for example 60 to 240 will work way better than 30 to 120 so its biggest flaw is not being that good for lower end hardware yet.

You say yet, but frame generation from a low base framerate would probably require a profound difference in how the frames are generated. When the base framerate is higher, I'm guessing there's two advantages over having a lower framerate:

• The wait between frames takes a shorter time, so interpolating new frames between the non-generated frames doesn't need to wait as long for the two frames to be available, so there's a major advantage in input latency.
• Because at higher framerates there's a shorter duration between individual frames, less change (e.g. characters/objects moving) takes place between two non-generated frames. Hence generating new frames in between is less likely to result in visual discrepancies.

Like I said, this seems like a profound limitation of the current frame generation technique. I'm sure there's tricks that can be used to improve the situation somewhat even at lower framerates, and of course it might be possible to come up with entirely new techniques, but it's more work than refining what you already have.

In best case it doesn’t matter since if the frames generated happens when frames for your input is ok and thus the game runs smooth enough anyway. Then it’s just a waste of resources.

In worst case it’s another excuse for horrible lazy developers to release unoptimized crap, that might look nice and smooth when you are watching it, but plays and controls horribly because the frames for input is low.

So all in all it just a waste of electricity.

I don't see much use cases for this where it would really matter.

If the frame rate is low, then fake frames will rise the FPS... But in that case the increased latency will make it feel much slower. In other words, the main use case for FG is moot.

The only reason I see to use it would be for the FPS to match high the refresh rate of a gaming monitor when it otherwise couldn't reach that number.

I do use to get to 120fps limit of my tv, if I get 70-80 fps without it

When reflex 2 finally comes out that'll probably help with the input latency when using it at a lower fps. It could take a while perhaps but it wouldn't surprise me if by the end of this decade it becomes good enough for to work well on lower end hardware.

I do since I play VR mostly and reprojecting 60 to 120 fps makes turning your head feel much better. GT7 improved on it on PS5 Pro, also feels sharper now rather than just smoother. Any improvements in making a 60fps game look like native 120fps are appreciated. (Dunno why you would need AI for that, motion vectors are a better solution)