[fe] Something like SpeedTree?

[Jason Weber]

On Fri, Mar 27, 2009 at 2:46 AM, Juhana Sadeharju
<Juhana.Sadeharju at uta.fi> wrote:
>
> Hello. How interested people would be in an open source
> tree software similar to SpeedTree?

I haven't used it, so I don't know the details of their system.
On current consumer hardware, I'm guessing it would be
strategically organized textures, perhaps a few hundred per tree
close up.  That would have to multi-res down to get any decent
quantity of trees to fill a background forest.  At some point in the
distance, they could revert to billboards or even texture walls for
forest edges.

I don't imagine the physics would go beyond simple lattice
operations you can do in a simple hardware shader.  I doubt
you will get real kinematics except may for a select few
situations.  They've probably concentrated on rendering,
so non-trivial dynamics might not even be in their toolset.

> Already there are some attempts to do so: greenhouse project,
> ngplant, opentreelib, arbaro, treal, and meshtree.
> Most of these are based on Weber's Siggraph 1995 paper.
> Also no realtime rendering is addressed that much despite wishes.

Weber and Penn.

> Ogre engine users have rendered trees in realtime but probably
> they don't have anything close to SpeedTree. Now Ogre users
> have possibility to license SpeedTree via OgreSpeedTree.
>
> SpeedTree license is unsuitable. I have asked SpeedTree people
> if they could release a cheap personal edition of their software:
> binary library, for personal use only, no redistribution rights,
> priced $50. The idea is that if people want run my application,
> they have to purchase the personal edition as well (compare to game
> modding). No, they could not release such edition.

If you run a studio, $10k to get instant trees for a game is probably
a good deal if the support is good.  You'll have to spend at least that
just integrating their code or 10 times that making your own solution.
But for open source development or recreational programming, any
non-commercial projects may as well disregard this as an option.
In any case, buying a solution also eliminates the learning process
and therein the resulting knowledge base in your staff (or self).

I don't really mind that a pay product is built on a free toolset,
but I do find it a bit inappropriate that they have pay product
as the feature link at the top of the Ogre page.  It makes you
question how free Ogre itself is or how long it will before they
try to go after people for money.  Oh, looking more deeply...
they're discussing "OUL license money", so I guess it is already
an issue.

> How Free Electron code should be modified for bringing it any
> closer to SpeedTree 5?
>
> Juhana

The demonstrated FE trees are a model for interactive kinematic response.
For general use, I would also want some robust multi-resolution physics.
You would also want a modern rendering platform, perhaps integrating
something like ogre, presuming there isn't some licensing problem there.

-- Jason Weber