CineMut shader family - Opaque

Post by **chuck_starchaser** » Tue Aug 19, 2008 10:54 pm

May not look like much progress, but the engine block is a very delicate thing; it has to be right for everything else to work, so it's taking time.

The good news is, once the engine is done, the rest is just tubes and hoses and wires; --trivial stuff.

vodalian · Post by **vodalian** » Wed Aug 20, 2008 3:48 am

that looks pretty awesome!

Post by **chuck_starchaser** » Wed Aug 20, 2008 7:47 am

Thanks!

Finished the left side of the engine block:

The right side is probably going to be more difficult. Not just because it appears to be more complex from what I can glean, but because it's hard to glean much. The right hand side of harleys are usually covered with the exhaust pipes. Anybody here has a Harley and a digital camera?

charlieg · Post by **charlieg** » Wed Aug 20, 2008 1:47 pm

Looks good!

Post by **chuck_starchaser** » Wed Aug 20, 2008 3:28 pm

Thanks! There should be a motorcycle riding adventure game, Charlieg. Maybe you could propose it a Freegamer. Why hasn't it been done yet? Just ride from town to town, watching the cows pass by, or trying to get past that slow 18-wheeler. At each major city there are a few biker hangouts, where you can drink beer, hear stories, maybe have some of your questions answered, solve a mystery, or get a new quest; and you can spend your money on drugs, or upgrading and maintaining your bike. And you can join a "faction" (not only motorcycle gangs, but also less shady motorcycling loving groups), or remain independent. There'd by wise motorcyclists you can also meet, who talk like Robert Pirsig ("Zen and the Art of Motorcycle Maintenance") for the enjoyment of philosopher players.
At the end of the game, you get a score that represents the number of quests solved, the final value of your bike and pocket cash, but you get big demerits for "dropping the bike", crashing, or for having to visit or pass by any city more than twice in order to solve the quests (so you'd take the blue highways whenever possible).
If any open source developers out there take on the idea, they get free bike models, CineMut shaders, some help with the physics, and a beer.

Post by **loki1950** » Wed Aug 20, 2008 3:42 pm

f any open source developers out there take on the idea, they get free bike models, shaders, and a beer.

And the beer would probably be the tipping point for most FOSS dev types

Enjoy the Choice

Post by **chuck_starchaser** » Thu Aug 21, 2008 6:27 pm

Well, I posted a formal proposal at Free Game Dev Forum,
http://forum.freegamedev.net/index.php? ... 5c4951fed8
but even with the beer thrown in, no takers so far

I've extended the crank-case to meet the cylinders, though they are not welded yet, and started to develop an outline for what will become the right-hand side cover plate.

Post by **klauss** » Thu Aug 21, 2008 6:53 pm

BTW chuck, I've been working with very-high-poly models lately and have a new piece of info for you.

Although GPUs can handle millions of polys per frame, they can't cope with that many per object. There seems to be a limit in vertex buffer size where performance drops drastically, and that limit is around 64k indices (vertices), and same number of triangles I imagine.

So you'll have to keep your single meshes smaller than that. Your object could be made of multiple <60k vert/poly meshes. Lower-end cards probably have a smaller limit, so I'd stay below, say, 20-30k?

You'll have a hard time reaching the 100k polies you wanted. Bummer right?

Post by **chuck_starchaser** » Thu Aug 21, 2008 7:03 pm

Very interesting! So, this probably means that vertex indexes are 16-bits wide, and when there's more than that, GPU's have to use some kind of paging scheme.
This is not "BAD news", necessarily. I was thinking about separating metals from dielectrics into two meshes, for different reasons. And the front fork has to be separate, so that if one day it needs to steer, it can. Plus the glassworks and lights would be yet another mesh. And each wheel should be separate. That would still leave me with some objects having more than 64 K vertices, but on the other hand, I have not tried to optimize anything at all. I've been cutting and cutting like Jack the Ripper; but once it's done, I'd probably be able to halve the vertex count; partly using scripts, partly by hand.

This is very useful new info. Thanks!

Post by **chuck_starchaser** » Thu Aug 21, 2008 8:59 pm

Hey; look what I found! A picture of the very motorcycle that the other model came from:
http://www.totalmotorcycle.com/photos/2 ... er1200.jpg
It was also a Harley.
I could have fixed the other model, instead...
EDIT:
Found a pic of the right hand side of the engine of an old Harley, and it seems the design hasn't changed much.
http://kathika.com/wp-content/uploads/harley.jpg

charlieg · Post by **charlieg** » Thu Aug 21, 2008 11:09 pm

Wow, the model and shader have come on tremendously! They look photo-realistic although when did you start rendering on white?

Aaaaaah

Seriously though, on the bright side, at least you have 2 bikes for your game once you have an interested programmer.

charlieg · Post by **charlieg** » Thu Aug 21, 2008 11:11 pm

In fact, the place you may want to ask for programmers is... the Ogre3d forums:
http://www.ogre3d.org/phpBB2

That may be a better place to solicit graphics / physics people.

Post by **chuck_starchaser** » Fri Aug 22, 2008 12:03 am

That's a good idea! Thanks! Just to clarify, I was just "throwing out" an idea; I don't want to get involved full time with it. My commitment is to PU first; then Tadpole. But I'll put a link to my proposal at the Ogre forum.

EDIT:
Done!
http://www.ogre3d.org/phpBB2/viewtopic. ... 652#301652

Post by **loki1950** » Sat Aug 23, 2008 2:14 am

Was checking out some links from that Model viewer i have been playing with and this showed it's self http://www-c.inria.fr/gamma/download/CYCLES/index0.php lots of other meshes in that data base

Enjoy the Choice

Post by **chuck_starchaser** » Sat Aug 23, 2008 4:06 pm

Great resource! Thanks! Low-poly, most of them, but it's great to have a starting point of a mesh.

The right hand side of the engine has driven me to depression. This is the third time I model that cover to end up with garbage; I show it not without shame, just to explain why it depresses me so much:

Finally, I decided I can't go on modeling blindly off pics in a magazine that show next to nothing, so I googled "harley engine" and clicked on Images, and voila:

Even a tatoo helps to clarify:

Shouldda done this long time ago.
So, deleting all the crap for the third time... I hope the fourth time it works, or I'll have to send the mesh to HOG and pay them to fix it.

Post by **klauss** » Sat Aug 23, 2008 8:59 pm

Why don't you try nurbs surface patches?
You can draw the contour first and then connect the lines, it's really cool and how things are done in CAD world.
Later you can convert it to a trimesh and refine it by hand.

You'll have to practice nurbs modelling first though.

Post by **chuck_starchaser** » Sat Aug 23, 2008 9:22 pm

Never done it; I have new-thing-fright to overcome, always.
When it comes to doing all the wires and cables, I'm gonna have to re-teach myself Blender's way of making spaghetti, and I'm already paranoid of that.
Thing is, the few times I used it (one curve serving as the path for another), it took me a full day of trying a million things to get it to work right; and in the end I never learned what exactly IS the right way to do it.

This has been going VERY slowly. Full day of work today, and haven't even got to the cover plate, yet. At least I think I'm on the right track, now.

Phlogios · Post by **Phlogios** » Sat Aug 23, 2008 9:26 pm

The only way I know to make wiring is making circles and using the spin function. It has been a long time since I read the wiring tutorial on... is it the PU forum?

Post by **chuck_starchaser** » Sat Aug 23, 2008 9:32 pm

There IS a tutorial at the PU forum, written by Dual Joe, but I couldn't understand any of it. He used a different way.
The method I was referring to is not by spinning a circle but py "pathing" it by another curve. You create a curve in space, and then in the "path" field you put the name of your circle curve, and press some button, and then you get the circle following that curve. Something like that. I did it two or 3 times; but each time took me like a whole day of experimentation to make it work.

Phlogios · Post by **Phlogios** » Sat Aug 23, 2008 9:34 pm

Which is exactly what DualJoe does, in the "oddly shaped hose" tutorial.

Post by **Deus Siddis** » Sun Aug 24, 2008 12:50 am

klauss wrote:Why don't you try nurbs surface patches?
You can draw the contour first and then connect the lines, it's really cool and how things are done in CAD world.
Later you can convert it to a trimesh and refine it by hand.
You'll have to practice nurbs modelling first though.

Don't let this discourage you all from experimenting with it, but as a small heads up, Blender is weak in the NURBS realm. Or at least it was last time I used it, but I have not heard of any changes to that part of the program since then. I only use it when I need specific control of the amount of polygons used in a moderately complex curvy surface, otherwise it's subsurf all the way.

I was never a huge NURBS fan, but it is a shame that it is one of the few weaker areas in the program.

Post by **chuck_starchaser** » Sun Aug 24, 2008 2:06 pm

I wrote:The right hand side of the engine has driven me to depression. This is the third time I model that cover to end up with garbage; I show it not without shame, just to explain why it depresses me so much:

Finally, I decided I can't go on modeling blindly off pics in a magazine that show next to nothing, so I googled "harley engine" and clicked on Images, and voila:

Even a tatoo helps to clarify:

Quoted my older post just to have the reference pics handy again.
Decided to start from the cover plate, see if doing so helps me figure out the engine block shape behind it. Still garbage, but not as bad as before.

Post by **chuck_starchaser** » Sun Aug 24, 2008 10:36 pm

Alright, front part of right hand side done.

Now, onto trying to understand the back section...

Post by **chuck_starchaser** » Tue Aug 26, 2008 5:27 am

Not much progress today, but, as I firmly believe is the way to get things done, I did something and not nothing

The problem with the right hand side of the engine continues to be its being so hard to understand from pics.

Phlogios · Post by **Phlogios** » Tue Aug 26, 2008 6:10 am

Maybe you could ask Maverick to take some pictures!