[FOLIAGE] ~ My fourth attempt at making trees, I'd use some critique and advice...
triangle
I'm pretty sure I could've done better, but I dont know how.
There's something wrong to it....but what? -.-'
Help, please... -.-
There's something wrong to it....but what? -.-'
Help, please... -.-
Replies
- Where are the roots?
- Your lighting is coming from above, so I can't see the bark/trunk of the tree. It just looks like a cylinder and the textures look black and grey.
- Trees grow upwards, they don't follow the angle of the ground like yours are (though that is just a placement issue).
Before going into any critique of your leaves or leaf cards, I'd want to see your references and the type of tree you're simulating. Are you sculpting trees in ZBrush?
- canopies have noticeable repetition (it's better to have a few different textures for them and mix them up)
- leaves scale might be a little big and therefore it looks a bit artificial
- leaves density is way too high. You might want to reduce it to bring more sun inside the woods area and more natural leafs shadow instead of this flat shadow that you have right now. It looks like the character is in the tunnel.
Good luck with your trees, keep posting your progress and say if something was helpful at all.
Like Spongebob so eloquently stated,
It's a joke, but I'm sure you get the point. Just a matter of learning more about trees.
My lighting is very basic, I just used a postFX volume to give colors a bit more of contrast, which is (I guess) the reason behind the dark color of the bark texture.
Roots are a useless extra, at this stage, since I'm just trying to achieve a good looking crown without spending too many polygons on leafs cards (although not even that has been properly tweaked, yet).
I dont really care about the bark, at the moment, what I want to achieve is a plausible maple tree. Not realistic, just plausible and kinda good looking.
The game such type of asset is going to be used in will have a fantasy setting and assets will look halfway between cartoonish and realistic....like realistic proportions but simplified (probably hand painted) textures.
The slight tilt (10° max) was just a placement test with the foliage tool, not really a final thing....like, at all.
And yes, I sculpted the tree in ZBrush. Actually, the base volume and the disposition of branches was blocked out in Blender, it's easier to handle....
And this is the alpha of the leafs (sculpted in ZBrush, ça va sans dire)....
EDIT: @BIGTIMEMASTER There you go. ^.^
make your eyes blurry, and look at your model, and then the reference. What are the general shapes you notice between the two? In your model, I see sparse, horizontal planes. Like a savanna tree (dont know the species name, but that iconic african tree). Whereas with the refrence, I see cloud-like, blobby clumps of leafs that get smaller andnoisier towards the top.
That is the biggest thing to my eye -- the general shape is wrong.
I'd leave lighting alone and focus on just getting that general shape correct, then you can fine tune your leaf cards, textures, and then finally focus on your final presentation.
Maybe you mean the Acacia tree?
Unfortunately, they were all about SpeedTree, obsolete methods, third choice free-wares, still rendering polycount budgets, awfully looking trees and, the most hated of all, poplars.....so many poplars, like people cant make trees of any other type!!
I want to make a tree with a large, full crown (as in the references), but without resorting to Fortnite's crappy "green bubbles".
Branches and leafs should all be there, so that the player can actually look upward and see through the tree's crown.
I might be a bit too ambitious, for a beginner, but I dont think settling for the easy way will get me anywhere. =0
The technical aspects -- that's the easy thing. All you got to do is explain in simple terms what you want to do, and either google or somebody here can teach you how. I believe you already have all the technical knowledge to pull this off already. The real work is in studying and understanding your subject. You can be using methods from 2001, but if you know your subject through and through, it's going to look good. So I would just really compare and contrast the reference and your work so far. What is similar, what is differeent. Using only what technicques you know right now, what can you do to get your model closer ot the reference?
Or am I morally allowed to? ^^'
How do I decide where 3D branches should stop and cards should start?
Right now, cards are relatively big and they cover for much of the lack of shunts in branches, they kind of hide the empty areas. I thought that was what I should've tried to do.....
Should I make more shunts in branches and reduce the size of cards?
You aren't going to get everything perfect this first time (or fifth). Just think of this as a training exercise. Go hog wild, and plan to work the entire thing over multiple times.
I can recommend you to try a free software called "Tree it". You don't have to use it for your work but instead you could try to measure your polycount while you are after a certain quality level. And you'll realise faster what you could expect to see in theory from your target budget. For example I was looking for a low poly trees that will look good and that was my result: (about 1500 triangles)
The problem comes with retopology. At the moment, no automated function is capable of achieving a good distribution of polygons within a reasonable budget. If I go too low smaller parts start to disappear, eaten away.
The (pseudo) low poly you see in the UE4 screenshots is a whopping 6k(!!) triangles just for the trunk+branches, where my target would be more like 3k and less, which is totally feasible (and already halfway done in Blender). Being it a full manual retopo, though, it's time consuming. =p
So...where do I stop with making branches, how much of that empty space between the current branches should I fill with new ones?
It's gonna be a third person game, players will walk through woods in an open-world/landscape environment, I cant settle for a 6 sided trunks with short 4 sided branches glued onto it and then covered with blurry cards.
Dont get me wrong, these trees look absolutely *great*, for a top down (RTS?) game with a limited zoom-in.
I'll look into Tree It, though, maybe I'll get some nice starting points for where cards should start taking meshes' place.
This is a case.
I'm testing it in the game engine.
I still dont have an answer for this case.
Thus I'm asking for advice.
Your advice sounds like "stop asking for advice, keep trying on your own....possibly until somebody comes over and tells you that you've been doing it wrong all along for the last two years".
I have a past freckled with self-taught methods and solutions (in many areas) that proved to be wrong or utterly infefficient because I was alone on my way. If all I can think of is the wrong way, how may I find the right one on my own?
This is like when you learn to spell a word the wrong way and nobody ever corrects you, so you end up spending 30-40 years with some faulty knowledge that will ultimately cripple your achievements.
I'd rather learn a better, more correct technique now, please.
Could you tell me how *YOU* would carry out this task?
How many sides would you make your trunk with? And branches? What issues may I encounter, depending on these parameters?
Would you sculpt it in ZBrush or just splat it with a tileable texture/normal?
Would you plug branches into the trunk as separate meshes or retopologize the tree as a whole piece around the ZBrush sculpt as I did?
How accurately would you follow the reference's shapes, to prevent the polycount from skyrocketing?
How many twigs would you bundle into one card, how big the card, what polycount/size ratio (roughly!) would you use, for a maple tree, to keep the polycount within a viable range?
Perhaps you need a tutor? Self-teaching requires the confidence to make judgement calls, I think you need somebody to help you develop the ability to do that. Probably won't take much, but perhaps one of these CGMA courses or somethihng similar would give you the boost you need to get past some of these hurdles so you can learn to make judgement calls on your own.
Or, actually, is there some shortcut/quick solution I didnt think of? =o
To give you an example of what I mean by that, I've heard of people who cook pasta by putting it in the pot when the water is still cold. They evidently guessed that method on their own. Maybe they can even achieve edible results, but then again wouldnt it have been better if someone told them since the very beginning that boiling water is where to start?
Or....imagine wandering in a 3D viewport without any grid, axis or even just a display with the current coordinates. How are you ever going to know what direction to take, if you decide you wanna move at 10 on the X axis? It could take you ages just to figure out if you're even moving along one single axis at all, let alone the right one by the right distance (and in the right direction!!).
So...yeah, a tutor....or someone who can point me toward a tutorial where they make a good looking tree with a large, lush crown, branches that are one piece with the trunk and leafs that are arranged into cards rather than bogged up into bubbly meshes.
Either one.....possibly the cheaper one.
It wont work for me.
What "Tree It" crafts are cones glued onto each other, I already do that much in Blender with the skin modifier. But then I bring it into ZBrush, dynamesh it into one piece and sculpt it so that it looks more natural in its shapes.
My mesh is a single piece, (hopefully) a more naturally looking arrangement of branches that depart from the main trunk like they actually sprouted out of it.
Anyway, is there a way to set branchlet's length separately from branches'? When I make branches longer, branchlets go crazy and the lower ones get so long they touch the ground! x)
It's hard to explain, better toch everything by yourself and see the feedback in viewport.
I don't think though that you need branches connected to the trunk nicely. That will increase amount of details and won't give you a good quality boost, because it's hard to notice.
You don't need a tutorial that shows you exactly how to do the exact thing you need. Techniques apply to all things; learn the technique, learn how to tweak it, and make it work for you and your needs. Experiment. I happen to know 100% that there are a bunch of great beginning-to-end foliage tutorials, using modern techniques, out there. And even if the tutorial involves speedtree, if it shows wires, and the results, it can still be dissected to learn quite a lot.
This videos are an absolute classic as well, full of amazing information. Though it's a paid tutorial.
http://3dmotive.com/series/creating-foliage-for-udk.html
Check out the trees here, look at how they're made:
https://80.lv/articles/a-free-tree-pack/
Check out Polycount's wiki on foliage:
http://wiki.polycount.com/wiki/Foliage
These articles with tree breakdowns/info:
https://80.lv/articles/jan-tverdik-cryengine-environment-design/
https://80.lv/articles/paris-diorama-the-power-of-details/
https://80.lv/articles/lessons-in-environment-art-production/
https://80.lv/articles/tips-on-creating-realistic-foliage/
Each of those articles may not tell you exactly how to achieve what you want, but they show some guidelines it's safe to work within. And though not all the trees look great, the techniques they use are solid, and should give you a good starting point, with all the other stuff people have mentioned.
As I mentioned in another thread, go into some games, and try to figure out how they did it, too. Again, you won't get the info handed to you by the game, but you'll get enough hints to set you going in the right direction.
These videos are absolute classics as well, full of amazing information. Though it's a paid tutorial, $19, and the game-engine specific info is for UDK, so a bit has changed since then. Still, incredibly useful.
http://3dmotive.com/series/creating-foliage-for-udk.html
And no, I have no game/model reference for this particular type of tree, and only a vague idea of a poly budget (3k for the trunk and maybe 2k for the leafs). Two things I've been asking for countless times in multiple threads.
Did you read all the replies prior to yours? A lot has been clarified about what you're asking.
But I didnt manage to fix branchlets, no matter what.
Could you try to make a tree that resembles the two references I've uploaded? So to tell me whether it's even possible for someone who already has some experience with this software.
Not a detailed one, just a tree that has the same shapes, proportions and volumes (and leafs density and distribution) as in the references.
Whenever you find yourself thinking, "but I can't do this because...." just stop, and try it. It's the only way. If you can't do that, might as well quit.
Second one offers models with 30k+ triangles (talking about the acacia). Is that an acceptable budget? If so, then I can just go full auto with ZBrush's retopo.
What I noticed is that it doesnt care about UV tiles at all, I guess because it makes use of tileable photographic textures. Also, again, branches are just plugged into the main trunk, everything's a separate piece (or even more than one). Transitions are quite noticeable, if you get close up.
I have the polycount wiki link on my desktop all the time, I've gone through the grass in CryEngine one which actually helped me A LOT to understand the base techniques behind grass.
Unfortunately, most of the tree tuts are offline, but two of them are still there and plausibly applicable.
One does more or less what I do (except for the ZBrush sculpt), which is branches that are one piece with the trunk. It's so outdated that it doesnt even use normals and the texture is just some random brown noise to hide the fact that the texture is not tileable.
The other one uses separate splines (VERY interesting method!!) and just moves lofted meshes' vertices around over the surface of the trunk to make branches look like they come out of it. It doesnt show how one's supposed to hide the transition between meshes when it comes to textures, so I'm not sure how to make use of it.
The 80 Level articles are very bland, none of them shows any real in-depth technique. I see useful advice, but I'll be able to make use of it only in the furture, when making every single asset wont be an overwhelming challenge anymore.
I'm starting to realize maybe I wasnt clear enough about my workflow and what I'm trying to achieve. I'll post an update as soon as I'm done replying to all the others...
I've already started watching it and the guy sounds very competent. My hopes are high, for this one.
I'm gonna post an update, now, because I have a strong feeling that I should've spent some time explaining my current workflow and what I'd like to achieve.
Anyway, those 10-20k tris for a tree sound reassuring, but it's baffling how I still cant find *official*, confirmed sources for numbers when it comes to polygon budgets. =0
I'd like to explain my current workflow and what I'm trying to achieve, since I keep receiving links to tutorials that clash with my goals.
Workflow:
.- block out of the tree in Blender, where I use the skin modifier to make multiple "hoses" as a base structure for the next step (very similar to the splines method)
-. import the blocky structure into ZBrush, smooth and dynamesh to make it a whole piece and then I start using the clay buildup brush to give the overall shapes a more natural flow by smoothing the junctions between the trunk and the branches
.- as a result, volumes change into a more chaotic series of deformations and a texture comes out of the sculpting, which is nice; a touch of a custom alpha resembling the bark will do for the normals
-. automatic retopology gives me debatable results, a good base for baking normals wont be anything lower than 6-7k triangles and up to 12k depending on the complexity of shapes (a manual retopology doesnt take half of that); still acceptable for testing purposes, so 6k it is
.- normals for the trunk baked in Substance Painter (for a quick review of UV seams, if necessary)
-. sculpt and paint canopies in ZBrush, export as FBX and then bake normals/color/alpha on cards with xNormal (or Substance Painter, doesnt make much difference, although xNormal is much faster to set and use)
.- cards meshes for canopies in Blender
-. at this point I have a low poly trunk+branches complete with normals (and eventually handmade textures made in Substance Painter by exploiting the normal map) and leafs cards
.- in Blender I place some vertices here and there over the tree's surface (as a separate object) where I want cards to appear and then apply a particle system which will generate as many instances as vertices; some manual tweaking of the cards assures correct direction and a better coverage of the overall volume
-. join meshes and export for UE4
.- material shaders for trunk and leafs in UE4
Done.
I like making my own textures (although nothing truly good still comes out) and I'd rather not use photographic ones, especially for the type of game we're trying to make.
Also, I want to avoid visible joints in the mesh, especially where tileable textures cannot achieve invisible transitions between meshes. Thus the ZBrush pass and the handmade details.
This is time for reviewing and some Q&A.
You have a possible solution for my issue? Let me ask about its details and eventually report for things that dont seem to fit my case.
You link me a tutorial? Dont get mad if I cant find there what I need.
Anyway some good knowledge is emerging, here, and I see people eager to discuss and go in depth.
I appreciate your contributions, although most of them are motivational advice and rants because I dont smile and kowtow.
I dont really wanna make you angry, I'm just one who doesnt give up.
https://www.evolved-software.com/board/viewforum.php?f=29&sid=a644f49162bbe81f5e80a3f2709c5f04
This was the tree I've made for a few minutes with default textures (4199 polygons)
I didn't went into much detailes about leaves placement but you can expect something similar to this at this polycount. Alhough you can change the actual leaves boards to the "branch of leaves" boards and exaggerate even more.
I can see how this model could be used in a top down game like a 3D RTS.
Nothing like the kind of shapes I was trying to achieve and that presented me those issues with the branchlets, unfortunately.
And yes, you're right, I've noticed several posts about bugs with textures in their forums. ^^'
Thanks again for your effort, this sample gives me a better understanding of the possibilities of this kind of automation.
THANK YOU!! ^0^
Time for some more review, please, I truly need feedback....
@BIGTIMEMASTER @Joopson @MizuhaMD @Flight @Ashervisalis ;
It took me a few days of awkward approaches, to devise a workflow that involved splines, lofting, ZBrush and retopology and that would give me a satisfying result with a *reasonable* amount of polygons (and in a reasonable amount of time), but I think I got it right.
This model counts about 5k triangles and the normal map is baked from a ZBrush highpoly, which is what I wanted.
Dont mind the quality of the bark or the colors, this is most of all an attempt to get the right proportions and a good distribution of leafs, along with natural looking connections between trunk and branches.
Now I have to craft branchlets and leafs, bake them into cards and eventually apply them to the branches.
It shouldnt take too long thanks to the video Shyralon linked me, which shows a very clever and efficient way to do it.
And now, let the critiques begin!! ^O^
Just keep in mind to set things up ahead of time so that, if it turns out your leaves cover a lot of your smaller branches to the point that they can't be seen, you can easily target them for removal.
I mean, don't rework things a thousand times while you are mostly focused on learning, but keep in mind that flexible approach in general. At any point in the process you don't want to be "stuck" with whatever you got at the moment.
As somebody else mentioned, having a few roots pop up from the surface a little is a nice touch. Not all trees will be like that, but it gives a sense of grounding.
It takes around 5k tris, but I think it could've worked just as well with 4-4.5k.
I kind of followed your advice to not care too much about the polycount and just go for accuracy. The tutorial Shyralon linked gave me the base technique, then I applied a thing I noticed in one of the screenshots from SpeedTree that you people showed me in one of my previous threads about foliage and eventually found a way to apply it manually in an efficient way.
Shyralon's video was THE thing I needed, though. Without that one I could've kept banging my head on this wall for the rest of eternity. =p
NOW I have an idea of how many polygons I should dedicate to each part, and where 3D should stop and cards should take over.