Aesthetics are an opinion, but it’s a fact that dynamic GI is not to a point where it can be more accurate than precomputed GI.
Frostbite GI is not fully dynamic, Cryengine is limited in quality and capability, the LPV system is super inaccurate and has lots of issues (which is why Epic isn’t messing with it much). I don’t know what Snowdrop uses, but it’s probably got similar issues because the new consoles don’t have enough power to do something better like what Epic was developing.
Read some articles about dynamic GI on google, in some way you are right, it’s not an opinion but the “opinions” or just call it experience go in two different ways, one site is claiming Full Dynamic Gi is much more better than precomputed light and the other site is claiming the opsite. You have also a lot of issues with lightmaps, like with every technology and i can just repeat and repeat mysel, if dynamic GI is so bad, why such a amount of games are using it? It is simply the more advanced technology and it will become standard sooner or later.
To create a world “game” which behavior and looks so close as possibible like the real world we are living in, full dynamic GI is definitly the future, shadows and lights are not precalculated in the real world and they are also flexible. Game engines try for years or with every new “generation” to get closer and closer to the “real life / world”.
For the power you are right, you need definitly much more power to use a full dynamic GI system. But for every more adavanced and realistic feature you will need more power! 
Just read this Unreal Engine 4 Supports Full Dynamic Lighting, Epic Working On Dynamic GI Solution | N4G
There is nothing more to say, i can imagine that Full Dynamic GI is a problem so far, maybe extremly bugy or what ever, so there is a reason why Epic released UE4 without Dynamic GI… but i’m sure it will come one day…
And for the snowdrop engine, just take a look and The Division, they are using full dynamic GI, and to be hornest it is so far the best looking game i have ever seen on a PC…
You keep repeating yourself but you are not listening to what others are saying. Nobody is denying that dynamic GI is the future, but none of the current solutions are perfect. Even SVOGI had problems and thats one of the reasons why Epic dumped it.
Btw, The Division is postponed, most probably because of scalability problems or other rendering issues they have to handle.
They have no choice–Battlefield 4 has such large maps that lightmaps would take up too much memory. Any games that need to adjust the time of day don’t have a choice to use lightmaps because the light is moving.
And that article doesn’t mean what you think it means, it’s written a bit poorly, but basically the visual differences we got for the Elemental demo from dynamic vs. precomputed GI is for things that are moving, like how the shadows from the doors aren’t correct because the doors need to animate, or how it’s lacking dynamic lighting from emissive textures. In any case, there aren’t any engines that offer that type of GI that was originally in UE4 anyways–Cryengine does not have that.
It’s all a matter of how much/what needs to animate, for the vast majority of games precomputed lightings is a viable option because the only thing moving around is the character and they have solutions to give a fake GI for that.
Of course i’m listening, the thing is… if someone is teling me, prerendered light is like the best solution so far “it’s the way it sounds to me” without any facts, it’s simply not plausible for me, it’s not the truth.
This is no offense, it’s just a dicussion… every gamer, coder, artist… etc… has his own special experience about this. Fact is, that a big group of UE4 users want the full Dynamic GI technology, cause the results you can get from it are simply amzing. That’s all i’m saying.
And if we compare the results with and without full dynamic GI, from my posted article, i think it is very good to understand what the better rsult was or is… that also didn’t mean, that this will work out for evey game, for every special demo or what ever…
And for the division, i know that the release was postponed but this didn’t change the fact the this game looks just amazing
@darthviper107 thats, what i’m saying all the time… it depends on the game typ and what your goal is…
however it was a nice duscussion i’m off for today…
Just to point out what you said:
You said Dynamic GI is more realistic, it’s not.
We can start it all over again but we should end this here, cause it just makes no sense
I’m sure that full dynamic GI provides much more realistic outdoor scenes and i just gave you a lot of samples / facts why it is so…
So anyway thats it for me…
in a static scene where nothing moves (or for just a screenshot), dynamic GI cannot beat baked GI simply because it has more power and time to process it. if the examples you’re seeing look better for dynamic GI you’re clearly looking at the wrong examples, or comparisons that don’t match (ie. different engines or just different scenes made by different artists).
if dynamic GI produced better results, all the offline 3d renderers like VRay or MentalRay would use dynamic GI…
Atmosphere Scattering is already implemented in UE4 with the Atmosphere Fog Actor, though I’m not sure which technique it uses to do the scattering.
I personally do NOT believe that Dynamic GI makes a better looking outdoor environment, there are many factors that contribute to it more than having a moving sun. For starters, go outside right now, and tell me if you can visually see the sun moving. Of course you can’t, because it’s so slow. Pick a good sun location and STICK WITH IT, unless you’re developing a game where 99% of the time is spent outdoors. The fascination with having altering time of day is really boring now and it’s rarely useful to be honest…
I actually find that a sense of human > world scale proportions is what makes an outdoor environment look good, that and very large very obvious shadows. Too many people try to make outdoor environments that make you feel closed in and without distant detail, and very rarely give the impression that they’re part of a much larger world. Skydomes that hit the horizon and tall cliffs at the edge of the map are usually the giveaways. It’s all about maintaining detail at a distance, Many Gears of War 3 levels achieved this (the DLC level ‘Nowhere’ in particular), along with even some UT3 levels. You can’t tell me that the art of level design of either series wasn’t up-to-scratch. The key is to make your player really feel like their in a world larger than the playable area.
I often find that levels with full DGI are usually very poorly put together purely to show off the fact that they have Dynamic GI, or you have to make significant cutbacks in the LOD of the level. The entire level then suffers. Battlefield 3 and 4 both also pull off the scale of the world better than most games, and they don’t have a TOD system, but their shadowing is probably some of the best I’ve seen…
TL;DR, TOD is overrated, as is DGI, next to good level design.
Full Dynamic GI would still not be more realistic, and no game engines use a full dynamic GI anyways.
Thank you for these thoughts and tips. As somebody interested in making the world itself the “main attraction,” these warnings are valuable.
Questions/curiosities: what if you aren’t going for realism, but you still want a system of dynamic light in an outdoor type environment? Maybe you want the world to feel alive. And alongside AI automatons, weather, static/dynamic architecture/landscape, you might also want a system where the light changes. Day/night is really just one method of this (obviously inspired by irl), but you could implement other styles (very not realistic styles). The point being, is there a place for this, especially if not realism? Do you have any thoughts on this? Seems you have experience and confidence on the topic, so I’m interested in what you have to say and share.
Note: dynamic GI may not actually be what’s needed to do what I am suggesting. I am still learning terms and tech about lighting.
Thanks.
Using just dynamic GI is of course not the best idea, it is, lets say a combination of both. But having the possibility to use full dynmaic lighting “even partly” in a outdoor game, makes everything much more realistic.
A impressive game scene isn’t just a picture like a vray render, you have moving parts, flying rockets, sequences where the sun moves etc… A good game like The Divison or Battlefield 4, isn’t just about static meshes, things are moving are animated etc…
Enlighten is a good example of “combineing” both technologys…
http://www.youtube.com/watch?v=JZO1LF305UE#t=47
http://www.geomerics.com/enlighten/
Games like Battlefield 4 or the Division would never look like they look if not using full dynamic lighting with other technologys.
So no matter what “fans” of lightmass or precomputed lighting says, full dynamic lighting in combination with other technologies is the future.
And please don’t get me wrong, a precomputed light in a static scene, with like not really much movement or not a lot of animations etc… is of course the better solution when it comes to the quality.
If you’re wanting to do an exterior environment with a moving light (Day/Night) then you can do something like a directional light + an ambient/skylight, or try out the new LPV system to add some GI effects. Again, it’s not going to give as good of results as precomputed lighting, but it’s the only option you have for that type of situation because the light is moving.
As for stylized looks, that depends on what you’re going for. You might not need to try and get a GI effect for some styles.
I just can hold on to my opinion / experience ,you wont get the same outddor qulaity with just precomputed light in modern games.
If the light is moving, you can’t use precomputed, if the light isn’t moving then you can and it will have better quality.
I’ve said this a million times now
i think i gave you more than enough examples why and in which situation fully dynamic lighting is the better choice, so i think there is nothing more to say for me… cheers!
Honestly if I make a game with Unreal Engine 4 it will have (almost) only indoor levels. UE4 looks amazing in these settings, but for outdoors it’s no good at the moment.
Actually even Unitys dynamic lighting for outdoor scenes looks better than UE4 and then ofc there is CRYENGINE for that.
However I’m sure pretty transition levels that have small outdoor sections could be created with UE4 precomputed lighting! The key here would be that they HAVE to stay small for Lightmass to work them properly.
I’d dare to say anyone who creates a sprawling outdoor scene and expects to use the lighting systems currently available in UE4 is going to have a bad time.
Thats absolutly right and thats what i trying to explain the whole time… UE4 is a amazing engine, no question but with the current lightmass it’s not a optimal solution for outdoor games.
Other engines like the cryengine “which uses full dynamic ligting” can deliver much more quality for outdoor scenes… but as i wrote i’m sure or hope UE4 will work on this…
What you’ve been saying is hardly consistent. The idea that dynamic GI is more realistic is just plain wrong, sure it’s the best choice in some situations, “outdoors” is not one of them, “outdoors with a moving sun” is one.
You can do it just fine for outdoor scenes, the things to remember that large outdoor scenes will take a lot of memory to render lightmass, which studios will definitely do but not all people have a computer with enough memory. The other thing is hard drive space, and a large outdoor scene would have a lot of disk space just for lightmaps.
It’s simply not right, lightmass isn’t the best choise for outdoor scenes “not for all”, modern games are using a lot of movement, animations etc… and you simply don’t have the flexibility to handle it with lightmass. The internet is full with articles and samples of which resolution is the best one for a modern outdoor game, which is rich in realistic, animations and and and… I think companys like EA “Battlefield 4” or Bohemia Interactive “DayZ, Arma3” and many more studios wouldn’t invest money and time for a dynamic GI solution when it isn’t worth it, that simple it is…
Memory, render time etc… is of course another problem of using lightmass…
But however, as i wrote i think at some point it is senseless to argue again and again… so cheers