Amazon have now free SpeedTree for Lumberyard (Cryengine). I dont know what deal they have with SpeedTree, but do you think Epic Games can make a likewise deal?
It would open up the barrier for everyone that might want to try out SpeedTree for their scenes.
It would be cool, but. If we had speed tree. I would’ve probably wasted 3-4 weeks messing with it before finally buying the nature packs I have. So. I think I’m better off as I’ve been known to fiddle with things that consume my time. Too many toys, and gadgets to play with causes----Oh look a butterfly.
Yes, we do have some SpeedTree veterans who I do think would find this very enjoyable.
It’s better you learn to create foliage manually than speedtree. Even if you are a coder. Trust me.
Epic doesn’t have the capital that Amazon has, Amazon paid allot of money to allow all of its developers to use ST8 for free for a year. Speed Tree doesn’t play around, they like to get paid. It would be exorbitantly expensive to do so, so expensive it doesn’t outweigh the benefits of ST, imo.
" SpeedTree® 8 for Lumberyard was released today under a free license provided by Amazon to all Lumberyard developers. This version of SpeedTree is free for Lumberyard developers, and includes updates over the next 12 months. "
So it will not be free for Lumberyard forever?
No it’s only free for 12 months.
Ah, thanks for the info 
Its probably better to stay with the subscription model then, which isnt that expensive.
It’s not only useful for very small teams. It’s standard tool for AAA games, especially open world games. You need lot of different foliage to create believable world.
Speedtree editor is designed for creating foliage. It’s lot faster to generate foliage using specialized workflow than reinventing wheel by doing everything manually in general purpose tools like Max. Usually you’re recreating real world foliage, so you want to use the same patterns for most of the games. And Speedtree utilizes these patterns.
Every decent engine supports Speedtree SDK which gives nice perfomance gain for foliage-specific meshes/materials.
I’m not artist and it’s very difficult for me to create even simple mesh in Max. Doing something like proper trees here is simply impossible without investing a lot of time to learn “traditional” modelling.
But it’s relatively easy to create foliage in Speedtree editor
[MENTION=3921]The Britain[/MENTION], I unfortunately don’t have time to spend an hour or two talking about all the details. And I’m not 100% firm on saying SpeedTree isn’t a good tool for -everyone-. CDProject did use it very nicely (though still not competitive quality), but majority of the audience, including the people here, aren’t going to be able to pull it off. The results you’ve got are probably consistent for you, but I don’t think I’ll be impressed if you post them here.
@kjustynski, Gives nice gain in performance? :)
Let’s import just one sample tree and see what happens.
http://uupload.ir/files/r479_12.jpg
1 Tree:
6 Materials (0 Material instances)
3 Texture sets
~8200 Tris
Where as if you tried to do the same thing manually it’d be more like
1 Tree:
1 Material (3 Material instances)
1-2 Texture set
~4000 Tris
Also, you’re probably aware that when you’re doing it manually, you can bake various stuff to your atlas i.e you can create 3 different trees from just 1 atlas compared to SpeedTree where it bakes a unique atlas -per asset-.
SpeedTree created atlas: http://uupload.ir/files/txaf_24.jpg
Manually created atlas: http://i.imgur.com/gxHhe.jpg
So doing it manually is still to this date the proper way to do it and all those giant studios such as EA aren’t re-inventing the wheel by hiring foliage artists to do it all manually for them, they know achieving the desired details require manual work. Besides, the freedom and control you have in something like Maya and Maya does not exist in SpeedTree. It’s good for small teams that don’t really want to bother with quality/performance, or are not looking forward to produce unique results (other than what everyone does with SpeedTree after downloading it).
Here’s a little old image of mine, (Left is manual work, Right is SpeedTree).
Overall, creating foliage manually takes longer, it has steeper learning curve as well, but if you learn that instead of learning SpeedTree, you will be able create what you have in mind and implement all the details you want, you’ll end up with something very optimized, and the end result is worth the effort. But if you’re not interested in any of that, then SpeedTree is for you.
Thanks for nice comparison. I’d to ask few more questions then 
Is it possible to easily compare in-game performance to? I see there’s more materials but SpeedTree provide some shader operations AFAIK? Isn’t there any gain from that? Or any other features of SpeedTree can’t be seen as seen object param?
And does 4k difference in tris is that important if we’re batching all foliage in… well… foliage system?
Also, I noticed that you used tree from free SpeedTree package. In theory it should well optimized mesh, what if it’s not? And toolset gives options to optimize the content?
Just asking a lot of questions because I was simply used to artists using (or wanted to use) SpeedTree. It’s tool that was only by bigger studios for many years. Now it’s easily available for small teams thanks to Unity and Unreal. So I assumed it should be quite optimized if it was used for many “prev gen console” games.
Well, maybe even CDP, Bethesda, Bioware, DICE (also part of EA) don’t have enough manpower to create foliage manually if they need create loads of plants. Is it possible that difference in actual game performance isn’t that huge?
Maybe just SpeedTree with Unreal don’t work that well? Which wouldn’t be surprising
That’d require to have 2 assets that are fairly the same i.e 2 apple trees and have the same amount of tri count. I don’t really have that. But since each material is a drawcall 1 vs 6 is a big difference.
Shader wise it has a wind effect (affected by wind actor), and AO baked into vertex colors. There’s no documentation on how to get wind actor affecting your custom made trees, but you can create your own wind setup and it’s fairly easy. Regarding vertex AO, it has it, but I generally don’t see any difference when I unplug/plug it. Oh and it has a SpeedTreeVariation node in material that randomizes the color on your instanced trees, but you don’t need SpeedTree since that node is already included in UE4 anyway and you can put it in any materials you want.
Well, if you have 1000 trees, that means 4,000,000 more tris (not considering the LODs) but at LOD0 that does make a noticeable performance difference.
The high tri count issue with speedtree originates from the leaf cards. The included atlases are very small, and when you generate a card around it, the card ends up being small too, so when the cards are small you need too many cards for the tree to look filled. Naturally more cards = More tris. To optimize that you’d have to reduce the number of cards, but that greatly hurts the look of it. Whereas if you did it manually, you could do bakes suitable for creating very large cards.
It depends. I’m sure when/if those studios use it, they don’t use it like we do. They probably create a different pipeline. I’m sure they don’t do “Generate>Export>Import>Repeat”.
All in all, in my personal opinion it’s not a tool you’d use for long time so I see no point in having it for free either. It also depends on the eyes. Some people see something as amazing and some others see the same thing as awful. You can try subscribing for a month and see how it goes for you. 
No you don’t.
That’s poor example. You just as well can end up with bad tree by modeling it manually, you just waste 10x times more time on doing it.
You can just as well create optimized tree using SpeedTree. You are just cutting yourself from very powerfull, specialized tool, which when used properly will always deliver better and quicker results than modeling by hand.
It’s like telling than modeling clothes in ZBrush+Maya is better than using Marvelous Designer. It’s just ridiculous. You will get better results in specialized tools, you just have to learn them first.
Either way SpeedTree 8 will be available for Unreal.
Hmmm that’s a wrong comparison. You can’t really compare Zbrush vs. Marvelous Designer to SpeedTree vs. Manual work. Just because a tool is designed to produce a certain type of assets it doesn’t mean it’s perfect nor superior compared to manual workflows. SpeedTree in particular works amazingly well for film industry where there’s virtually no limitations for artists. On the other hand, when it comes to games and optimization becomes a thing it works really inefficiently. Still, my personal opinion as someone who has tried different workflows. You’re free to believe what you want.
With so many options available I do not think Speed Tree is a fundamental Middleware solution nowadays.
They will never provide the tool for free, everything they do is promotional.
You will never have a free Speed Tree for ever. This will always be for a limited time ever.
But with so many wonderful procedural vegetation assets in the marketplace using only blueprint.
I think speed tree is becoming increasingly worthy of better promotions to attract buyers’ attention.
And there are many other options without subscription like the plantfactorie.
Anyway, the only interesting thing in speed tree is its integration with the Unreal Engine LOD system using billboards sprite sheets.
But nothing impossible for a probable procedural tree plugin that can be fully integrated within Unreal Editor.
If Amazon really wanted to increase interest in their engine by getting version 8 first and offering the modeller for free, perhaps they should have gone one more step and arranged to have at least 1 free Speedtree 8 model available. Because as far as I can tell, with me not wishing to try to create a Speedtree from scratch at this point, but just see how a version 8 Speedtree looks in Lumberyard, I cant actually do so without having to buy a model right now. I much preferred what happened with the previous version of Speedtree for Unity & UE4, where there was at least enough free content to be able to drop something into the editor and see how it actually looked before spending money. And I certainly did spend money on both modeller subscriptions and tree packs. But only because I could see for myself the sort of results possible in game without spending up front first.