A Unity ID allows you to buy and/or subscribe to Unity products and services, shop in the Asset Store and participate
in the Unity community.
Introducing the new Universal Render Pipeline and High Definition Render Pipeline subforums!
Unity 2019.3 Beta is out now.
Discussion in 'Made With Unity' started by BHS, Apr 16, 2011.
Where's this tool?
@Duke it's called profiler and it's only available in pro
RealTrees now available! Many more trees to come. Available on the 3dstudio.com (coming very soon to the Asset Store)
RealTrees Forest Pack 1
You know what would be amazing? If Unity Tech would buy licence for SpeedTree and include it in Pro version, just like they did with Umbra.
Actually if you ask me SpeedTree utility with engine is 10 times more usefull for development than Umbra. I am not saying Umbra isnt good, just i think that at least 90% of RPG/MMO/FPS developers would be much more satisfied with SpeedTree than Umbra.
OMG. You really dont know what are you speaking about. Umbra (or any oclussion system) is key to AAA graphics. Without occlussion, you are fu****, you would never be able to do large world with high poly models (or any more complex game on any platform). Some tree app is nice, but nice trees can be done in thousands off apps and even any 3D modelling program, any competent artist can do that. We need working Umbra (no crashes, multicore baking.....), no fancy tree modeller.
you shouldn't use OMG if you list Umbra in a sentence with a thing its not usable at all at the time.
Umbra only works with the main scene you load into, you can't stream it in in any form. so basically its limited to complex and dense scenes, in a big world you are not really able to use it at all due to not streaming and due to 2GB RAM just not being enough to even bake it
LOL, this is true. Anyway I was speaking about value of occlusion vs some fancy tree toy. I am waiting for some usable version of Umbra too, it would by really silly to use some custom occlusion system with such technology inside of Unity (btw, m2h occlusion looks nice).
The value is project specific. There are games that don't have large fancy worlds but need dense forest where speedtree would beat any benefit umbra would give and cases of Quake levels, canyons or curved city roads / mountain roads where umbra will beat any benefit speedtree would give.
in the end they both have value, but the tree system definitely already works quite a bit better than umbra so the current focus hopefully is umbra, not the tree system
Agree. But Umbra works on terrain too, so I can have city behind mountain and benefit from occlusion. I made some experiments and it really works, house behind hill is not rendered. Large flat terrain is very unprobable, so even large dense forest can make benefit from it (I dont know if tree system works with Umbra, anyway any custom mesh tree system should).
There can be only one improvement to tree system. Allowing to use additional LOD. So instead of having main model + automatic billboard, I would have main tree for close distance, one LOD for middle and billboard for far. Such modification should allow to increase polycount dramatically without bigger performace loose.
agree. being able to use the advanced tree creator shaders on custom made meshes would be another point.
Let me put it that way. Put up poll with: What would you rather see in Unity- SpeedTree or Umbra? Something is telling me that community would vote for ST rather than Umbra. It is dependent on project, but the fact is that most of projects would rather consider ST than Umbra.
Author of this topic showed, that even with integrated Unity tree system is possible to make very good trees. Quality of output is not only about tool, but about user. I dont know what is typical project for you, but for me it means finished game, with reasonable framerate on target platform. Actaualy, how many people here ever finished game? Votes from kids, who are not able to do anything and just begging for magic "create MMO button" have no value. Any serious game developer understands, that first we need stability and basic tools without bugs, than we can start adding new fancy features. Occlusion technology (in any form) is must for modern game engine.
I've been messing around with the tree creator some more and I've really gotten the hang of it. I find it a lot easier to manipulate the branches and leaves to how I want instead of it feeling kind of random like before.
I thought I'd try to do a pine tree so I did. Here are the results so far some feedback would be great. I did use mesh because I don't think billboarding would work with this type of tree.
Pine Tree 2,800 tris exactly
It looks very nice. It should fit in some game maybe with mix of cartoon and realism. However for proper real pine tree it is missing some things. Needles are too big. You should make smaller leafs or either scale up texture and make more needles in 1 texture. Right now looks good, but only for a very short tree.
@janpec Thanks. yeah I know the needles are big I'm trying to edit the textures.
Here's another tree this ones a lot more complex settings wise because I tweaked everything by hand but it gave me a great full looking tree with low tris of only 1920. This tree can also cast and receive full clear shadows. I did use some SpeedTree texture I found but I'm in the process of making my own.
Ik Animation - I think you're making amazing progress. I really hope you go on to get the maximum potential out of this
Ik Animation> Here's another tree this ones a lot more complex settings wise because I tweaked everything by hand but it gave me a great full looking tree with low tris of only 1920.
You on the right way, because OCC (Umbra) is not fine with custom multipoligonal tries at all. If you have those tries in scene, this is the best way to crash umbra or got incostent PVS data with blind areas near to inner parts of custom tree. Correct me if I wrong, but with new procedural Unity tree this problem does not occurs (I cant did this). On other hand, optimised good-looking trees are more difficult to perform is a case of procedural trees.
P.S. If you'll make 10-15 unique trees, simmilar with you upper examples, I will gladly to buy it on AssetStore for not much expensive price
If you don't want OC to crash, you better never mark trees as static (and not the terrain in consequence) cause it will generate a cell division to a degree where a single tree needs as much power as a whole 100x100x100 area with "non trees" needs. and that although it does not need it at all as trees consist mainly of transparent and transparent must not be marked static to work fine anyway.
You right, no other better way.
Static OC best only for big convex objects like city buildings, which really may hide something behind of them or scenes with room environment. Big forest more difficult to make with OC, but there is a trick: on each tree replace whole mesh to simple box (bonds to trunk), than compute OC, than replace box to actual trunk. As far as object_id not changed, this way allows to got smart occluding areas.