Can you tell which game engine is being used just from a screenshot?

I disagree. We don't.

A good idea would be to transfer well-tuned scene from unreal 4 to unity and see how the look changes. Finding "where EXACTLY the difference lies", IMO, is a meaningless pursuit. What matters is what you can do in the engine, not where they differ. Seeking "exact" difference, IMO, is being distracted/sidetracked.

I'm pretty sure this one isn't used at all by anyone. People use approximation, and there are a lot of those.

The whole idea of PBR workflow was that parameters are more or less the same. 0 smoothneess - completely non-smooth 1.0 smothness - completely smooth, 0 metallic - completely non-metal 1 metallic - completely metal. However, the details of the look depends on the equation used by the engine. That's why specular highlights in unity will look different compared to unreal 4.

In the end, I have a strong impression that it would be a good idea for you to got some practice writing shaders. Write a basic per-pixel phong for unity, with shadow and pixel light support, for example.Then throw in different specular model, then maybe try PBR approximation.

You can do that. You can throw out entire unity standard shader and replace it with your own setup. It is actually very easy to do in unity (and very hard to do in Unreal 4). Now, you can't easily change the way shadows work, but with a bit of effort you could perform rendering by yourself from a script. All of it.

 

Yeah that definitely looks like it may be the answer, I hope that it will be possible to use it to create graphics as good as other engines. But the fact that it is a response to the difficulty of creating a one-size-fits-all solution is a good sign, since I think this is probably what Unity has struggled with.

That's exactly what I said, straight after the part you quoted.

Well then, we need to find out what's the difference between Unity and UE, if we consider UE to be a benchmark of what Unity's graphics should be able to look like?

Although I'm very interested in computer graphics, and I'm practicing quite a bit with general shaders and compute shaders for various things I need, I don't think it should be necessary for me to learn/create custom lighting models and shaders just to make Unity graphically competitive with other engines.

But anyway, I probably will, because I enjoy it - luckily for me.

I've never doubted that it's possible for me to create my own illumination system for Unity if I wanted to - I mean, there's SEGI, and all that ... but I don't think it's something that should be necessary.

Anyway, I don't think I've been paying enough attention to scriptable render loops, which seems to be a clear acknowledgement of this problem and a step toward the solution that I like - namely separating high-end/low-end rendering.

I will probably download 5.6 soon and play around with it.

 

-_-
This should be possible in any decent game engine. Because non-photorealistic rendering, edge cases and stuff. Valve lab renderer (didn't have a good look at that one) provided custom renderer.

I strongly disagree.

If you want to utilize shaders to the fullest, you MUST know at least basic lighting models and must be able to implement them. This stuff is not that difficult to begin with, and if you skip it, then I'm not sure what the hell was the point in trying to "learn" shaders to begin with (or what is there left to learn).

It is pretty much one of the first subjects you would learn when studying shaders. I mean, phong lighting model is something you should be able to recall at a drop of the hat without any reference material. It is the basic stuff.

 

It's not difficult because I'm a programmer, but if I wasn't it would be a different story. And anyway, UE doesn't require anyone to program lighting models just to make stuff look good.

I have a million things that I want to work on, most of which are not core engine features, but which depend on them. So it would be great if I didn't have to 'roll my own' to get what I wanted, and what is already readily available in other engines.

Anyway, I'm going to learn the API for these scriptable render loops and see what can be done with them, since I think this is probably the answer I'm looking for. So I rest my case!!*
*for now