Forest Lighting study
polycounter lvl 7
Hey folks, I just wrapped up my most recent lighting study in UE4 and wanted to share how it turned out. Been building it up and testing some techniques for a while and while there is more to come this version is more or less done (seasons and weather coming soon though).
Updated with the newest version of the forest
Hey guys. Been a week or two but I am pleased to share the most recent version of the forest. It's not so much a new time of day or weather but I wanted to really push for something difficult. As of right now I'll be taking a few days break then getting to a breakdown that shows as much as I can.
With this one a video is probably a bit better to watch than looking at pictures. Also I had a massive night mare dealing with compression so for those curious here is a much nicer version of the video at around ~70mb (I highly recommend this link over youtube)
Direct link: https://www.dropbox.com/s/wdyc08wggbzgjzf/ForestFire_Optimized2.mp4?dl=0
[ame="
https://www.youtube.com/watch?v=--rMr4NiWUM"]https://www.youtube.com/watch?v=--rMr4NiWUM[/ame]
Update! People wanted a video so I spent some time and got something done quickly. Next update Ill add some textures and stuff
[vv]121312302[/vv]
Update 2! Fall pictures!
Update #2! Breakdown added at the end! I couldn't fit it all in here!
People asked so delivering at least part 1 of a breakdown for the forest day. Went into quite a bit of detail so I wont add much in text but I will add some. I will definitely be doing a 2nd breakdown of the forest fire but that will take quite a bit more work and I don't quite have time at the moment. Will be at least a few weeks for that one.
Lets start with some textures. For the most part, terrain and tree/rock textures were all built with displacement via tessellation and vert blending in mind. While not every texture had a proper roughness map, most were approximated using in game values. For reasons I'll go into later on other versions of this forest, I was insanely limited on texture space so every material had to fit a large amount of masks in only 3 textures. I also discovered that height for blending did NOT make good height for displacement and had to separate those 2 out. Every material except for foliage had an option to use world space blended textures instead of traditional UVs and could be tweaked, tinted, and adjusted in quite a few different ways.
Below are the main textures I used for terrain and rocks. For rocks specifically I went with low detail of large shapes because I knew I would use geometrey for form definition and vert painting between material types to create variation. I tried to stick with photosource for most of these but there was quite a bit of hand work done in places to keep the visual fidelity high.
Finally my end game takeaways from my terrain and rock texturework:
-I loved some of the detail that came from the photosourcing visually
-When it worked correctly the tessellation was a huge boon to the final visual quality of everything
-Tessellation just doesn't matter that much at a distance. Cant rely on it for ground breakup the same way you would up close without destroying performance
-Base geometrey detail matters for tessellation. I had to tessellate my ground plane VERY specifically to get the detail I wanted in the places I wanted
-Vert blending was as always an instant win for ease of use and variety
-For tessellation small shapes pretty much don't matter, but height variation for stuff like dirt is useful to add interest to the ground plane
-Photosource normals won't ever be as good as zbrushed or hand made textures without lots of custom work but it can be done
Up next shrubs! Not much to say here. A mix of photo source and using custom geo either rendered in Maya or SpeedTree to get good base textures. The grass was a ton of work to get to where it was but I would have loved it to be more photo realistic.
And now trees and tree bark. The bark had the option of vert painting 1 extra material and I took advantage of that specifically with the birch trees. The trees took a ton of R&D and I'm still super pleased with the beech trees. That being said the images explain themselves so I'll keep this part short.
Finally the foliage rendering. Below is a comparison as I added features and tweaked lighting on a single tree. Not everything I did was required for final image quality and while I used realtime LPV's in unreal, I got similar results with other techniques if performance is a potential problem.
Anyway, that about wraps it up for part 1. I'll be around for questions and a link the imgur gallery is here: http://imgur.com/a/fHBNK
I learned a metric ton doing this and hopefully someone else will too. On to the next project!
Updated with the newest version of the forest
Hey guys. Been a week or two but I am pleased to share the most recent version of the forest. It's not so much a new time of day or weather but I wanted to really push for something difficult. As of right now I'll be taking a few days break then getting to a breakdown that shows as much as I can.
With this one a video is probably a bit better to watch than looking at pictures. Also I had a massive night mare dealing with compression so for those curious here is a much nicer version of the video at around ~70mb (I highly recommend this link over youtube)
Direct link: https://www.dropbox.com/s/wdyc08wggbzgjzf/ForestFire_Optimized2.mp4?dl=0
[ame="
https://www.youtube.com/watch?v=--rMr4NiWUM"]https://www.youtube.com/watch?v=--rMr4NiWUM[/ame]Update! People wanted a video so I spent some time and got something done quickly. Next update Ill add some textures and stuff
[vv]121312302[/vv]
Update 2! Fall pictures!
Update #2! Breakdown added at the end! I couldn't fit it all in here!
People asked so delivering at least part 1 of a breakdown for the forest day. Went into quite a bit of detail so I wont add much in text but I will add some. I will definitely be doing a 2nd breakdown of the forest fire but that will take quite a bit more work and I don't quite have time at the moment. Will be at least a few weeks for that one.
Lets start with some textures. For the most part, terrain and tree/rock textures were all built with displacement via tessellation and vert blending in mind. While not every texture had a proper roughness map, most were approximated using in game values. For reasons I'll go into later on other versions of this forest, I was insanely limited on texture space so every material had to fit a large amount of masks in only 3 textures. I also discovered that height for blending did NOT make good height for displacement and had to separate those 2 out. Every material except for foliage had an option to use world space blended textures instead of traditional UVs and could be tweaked, tinted, and adjusted in quite a few different ways.
Below are the main textures I used for terrain and rocks. For rocks specifically I went with low detail of large shapes because I knew I would use geometrey for form definition and vert painting between material types to create variation. I tried to stick with photosource for most of these but there was quite a bit of hand work done in places to keep the visual fidelity high.
Finally my end game takeaways from my terrain and rock texturework:
-I loved some of the detail that came from the photosourcing visually
-When it worked correctly the tessellation was a huge boon to the final visual quality of everything
-Tessellation just doesn't matter that much at a distance. Cant rely on it for ground breakup the same way you would up close without destroying performance
-Base geometrey detail matters for tessellation. I had to tessellate my ground plane VERY specifically to get the detail I wanted in the places I wanted
-Vert blending was as always an instant win for ease of use and variety
-For tessellation small shapes pretty much don't matter, but height variation for stuff like dirt is useful to add interest to the ground plane
-Photosource normals won't ever be as good as zbrushed or hand made textures without lots of custom work but it can be done
Up next shrubs! Not much to say here. A mix of photo source and using custom geo either rendered in Maya or SpeedTree to get good base textures. The grass was a ton of work to get to where it was but I would have loved it to be more photo realistic.
And now trees and tree bark. The bark had the option of vert painting 1 extra material and I took advantage of that specifically with the birch trees. The trees took a ton of R&D and I'm still super pleased with the beech trees. That being said the images explain themselves so I'll keep this part short.
Finally the foliage rendering. Below is a comparison as I added features and tweaked lighting on a single tree. Not everything I did was required for final image quality and while I used realtime LPV's in unreal, I got similar results with other techniques if performance is a potential problem.
Anyway, that about wraps it up for part 1. I'll be around for questions and a link the imgur gallery is here: http://imgur.com/a/fHBNK
I learned a metric ton doing this and hopefully someone else will too. On to the next project!
Replies
I really like how you outlined the leaves on the second picture
There will definitely be a video of sorts soon but not quite yet. Still working on some other things first
Are you using the landscape tool or a static mesh? Also is your ground taking advantage of Dx11 or bump offsets? The textures look great and i'm struggling getting Dx11 working nicely on landscapes for a short anim and I'm contemplating switching to static mesh to avoid the problem.
I chose to go with static meshes for the terrain so that I could have better control over my geometry closer to the camera (and because I was iterating quite a bit). I used tessellation for close up detail on the ground. I did however use a ton of instanced foliage through the entire level. It was definitely a life saver on time.
EDIT: I see the caption at the bottom of the images now
Not yet, though I will post at least some render breakdowns and some textures in the coming days.
The entire scene is lit dynamically. I experimented with some light maps but they were not as foliage friendly when I started and I needed the ability to change light angles.
The trees were a mix of all kinds of techniques. The original leaf cards were made with renders of 3D branches which I then tested in speedtree to see what kind of coverage I would get with different techniques. I found that procedurally placing them caused too many weird dark spots due to light not traveling properly through super dense areas of leaf (which is to be expected because even with ambient light, and foliage or SSS rendering, it is still a game engine). I ended up doing most of the placement by hand, sometimes fudging accuracy quite a bit in the name of final visibility and lighting quality from ground level.
More soon,
Alex
Thanks! 100% of the foliage is 3d baked to planes.
Alex
[vv]121312302[/vv]
I'm a novice to UE4, I've only used it to present; and not very well at that.
Did you model this scene and light it? or did you get these assets out of a kit then light it?
I know since its a lighting study, thats probably what you were focusing on. Sorry if this question is a little confusing, I'm not even sure I'm asking the right question ! lol Thanks!
While you are correct that this was primarily a lighting study I did indeed make the models and textures as well (with a few minor additions I'll get to when I post a bit of a breakdown).
Yup those are all done with the light shaft bloom. Nothing fancy there.
Also updating the top post with new fall pictures too (though no video for this one)
I was building an open world in 4.5 and 4.6 but I got frustrated with the foliage lighting using the foliage tool.
I'm excited now to get started again if this problem is solved.
Yeah for the daytime It's definitely a valid point. I've been playing with the ambient light values back and forth pretty much constantly. Definitely hard to find a balance between close up and far back values since high intensity fill lighting up close washes out parts of the background and vice versa. The video definitely doesn't help. I may be getting better at natural lighting but damned if I can't figure out the inner mysteries of video compression
I know your video is a demo, and no where intended to be a game scene (yet) but I would like to point out a few things, leaves do not constantly fall from trees, the grass and foliage do not constantly move, wind flows through and around objects making it procedurally generated by nature, so this should reflect in the movements of everything.
I want everyone that is trying to replicate nature in the the digital world to do something, go to your nearest forest find a nice spot in the middle of everything and just sit there, maybe pack a lunch and have a picnic and just observe the flow of nature a notice how it is completely random in everything it does, the person that can develop this in the digital world would be a god among us all.
Imagine a game, let's say a survival game for which you are running through the a forest that has not been created by an artist but a procedural algorithm and rules given to the computer, I mean I wish I had a room full of programmers and developers so I can direct them somehow and create what I am thinking in my head, now keep in mind this would not be generated on the fly while playing the game.. that would be crazy hard on the CPU, GPU etc.. and it wouldn't really work on a multiplayer level, but once you have a base that can be randomly played-back depending on the situation the player runs into while playing the game, wow I'm just blowing my mind away right now.. and I know I'm not the first person to think this, but I registered on this forum to post this now before I forget what I was thinking.
Anyways I'm just rambling on now..
Kind Regards,
-b3ck
Can you give performance details ? What is your system and what is the average FPS ? Of course this is after the build and packaging right ?
Keep it up. Subscribed to this thread
Absolutely incredible.
Thanks for the feedback and the patience!
Alex
Thx