Tutorial: Realistic Medium Poly Tree
Bryan Tenorio presents a detailed tutorial on tree creation in Blender
Vegetation has always been a neglected topic when it comes to tutorials. So I have decided to give the community exactly what they have been asking for. This method makes a tree with a reasonable number of faces. (mine came out to 7,128) This allows for having multiple trees without your scene becoming too heavy. Another great thing about this method, is that you can use it in the blender game engine. If you’re not using Blender, you should be able to use the same methods in whichever program you are using.
In this tutorial you will be learning
- Modeling from details first
- UV Unwrapping
- Creating normal Maps using xNormal
- Applying Textures
Link
Haven’t watched this, but should be good I guess.
quite simple but smart technique! the result is awsome and it has enough polygons for almost every purpose.
7K faces for a tree is too much for a game asset though. o.O Maybe as part of a visualisation demo or game proof-of-concept.
The tutorial is quite easy to follow, and the end result is impressive. It should be noted that the normal maps can be created within blender too, through baking ;)
Any special reason why did you use Xnormal program instead baking the textures within Blender?
@Parallax
I should have clarified. 7K is too much to populate a level with, but you can use the tree as sort of a hero object, where there’s only one or 2. The preview image is actually taken from the game engine.
@Artorp and Animaticoides
I know I could have baked the textures in Blender, (that was the original plan) however it was being really buggy with me at the moment of writing. I wanted to use Xnormal to convert the bark hightmap into a normal map anyways, so I thought I might as well show how to do it in there as well.
Excellent, thank you Bryan.
Clear, light and simply well executed!
Thumb up! you didn’t try to become a youtubee star. :)
I’m just curious why Xnormal is used… you can create the bark normal map with GIMP filter NormalMap (duh :P) in 3 mouse clicks and 1 second, keeping all the texture work just in GIMP without the need to import/export to a third app. Besides baking render passes with Blender is delightfully easy.
nice tree model anyway ;)
Anyone knows Arboro (for Povray)
It’s an Java based tree generator, at least I thought.
here’s the link
http://arbaro.sourceforge.net/
Does anyone knows it and does anyone know how to get the textures right?
Grtz
Nice! Just what I was looking for! :D
Thanks, Bryan! (Boy, it’s always weird addressing another Brian!) I take a good Blender tutorial any day!
Just an interesting side-note with regards to creating believable organic shapes and geometry.
I just heard of Unlimited Detail http://unlimiteddetailtechnology.com/ and had a quick look through their videos. Apart from pretty horrible background noise/music, their voxel(?)-based graphics solution seems very impressive since they claim to run huge point-cloud based scenes in real-time on CPU.
Especially, detailed shapes like leaves, branches, roots and other organic shapes, seem to possibly benefit from this type of rendering.
Will be exciting to see if this turns out well in the coming months…
@Jimmy
that website was pretty surreal ;D
too bad theres no real explanation how they achieve ‘unlimited’, guess they won’t let anybody steal their idea…
Apart from the strange music and the constant repetition of the term ‘unlimited point cloud data’ that site was pretty interesting…hard to believe but still awesome ….
On the other hand this just has to bring up skepticism since it seems too good to be true, and how would a comp manage unlimited point clouds at all..esp. since they mentioned nintendo ds as an example of a handheld console to process it. I browsed through the ‘what is it’ section, but the brief analogy to a search engine didn’t really clear things up for me. Spooky site:D
Definetly a very odd and must see link imo
Need more info, totally amazed and confused about this system. Trying to imagine how a development kit for that would look like.
Thanks for sharing.
I forgot:
Cool tree tutorial by the way!!! :)
addendum to my previous post:
Now that i tried to look deeper into the claims made by unlimited detail, i gotta admit that I came to the conclusion that this technique is far from being an alternative to poly based engines.
Theres no animation/ interactivity whatsoever, their point cloud objects are often repeated thousands of times on a grid and so forth.
I can’t shake of the feeling that this is just another example of vaporware/ bloatware that is most unlikely to ever happen anytime soon.
Anyhow gave me the chance again to think about polygons and that i should be careful with my enthusiam towards the claims made there;D
Still I’m eager to say if this ever becomes a practical reality.
Well, actually, they also claim to support animation. http://www.youtube.com/watch?v=cF8A4bsfKH8 . There is also similar work on voxel rendering being done by nVidia, so they are not alone in following this vein of research.
Thinking about it: Imagine a very detailed 3D model of a rock. There is no bump-map, normal-map, parallax occlusion or other screen-space processing in effect. Given the amount of geometry detail (number of polygons), when rendering the visible polygons of this rock, each could maybe be so small as to take up only 1px of screen-space. Even so, the polygon still has to be rotated by the GPU to fit the current view, the texture on it has to be stretched to fit the rotated shape etc, etc. All this work ends up as only one pixel on screen.
Now, imagine you’re rendering a point cloud: All that has to be done for a give screen pixel is to “shoot a ray from the camera” until it hits the first point in the cloud, check the color of that cloud point and assign it to the screen space pixel. There is no need to do work on rotating or stretching a texture to fit onto a screen-space polygon. However, in the cases where the points are too far apart (i.e. camera too close), some screen-space filling-in would have to take place, I assume.
Given the example of the detailed 3D rock model, it’s possible to imagine that the number of points in the cloud is the same as the number of vertices in the regular polygon-based model. Now, whether it is less work to render the rock as voxels or as polygons, is still an open question to me, but I’m very excited to see that someone are trying to figure it out.
And yes: Thanks for the tree tutorial. I’ve always been reluctant to try since I’ve always ended up with som kind of cardboard-looking tree, but now I know how it’s supposed to be done :)
There is a problem with z-transparent faces they don’t support shadows