Some inspiration

[charlieg]

http://www.gianlucamiragoli.com/immagini.html

hock: hock: hock: hock: hock: hock: hock: hock:

Sadly I think the models are owned by his employer.

[charlieg]

Yeah, damn, for this game: http://www.testdriveunlimited.com/

[zimluura]

i was wondering if that would be a good way to do the lines between the body panels...
more mesh detail and less texturing work.

edit: rough wip on the tc6

[joevenzon]

Hmm, that looks quite nice.

[thelusiv]

zimluura, the TC6 looks great. I can't wait to see it finished, it will look very nice.

[zimluura]

thanks! glad you like my obsession!
been rethinking a few things about how to design these meshes.

with the gaps it's easier to select chunks of a single object in blender with the ctrl+L (select linked) key. so that way i could easily split the car into it's components later.

and after thinking about that for a little while i threw in some low-poly placeholder meshes for the engine, intercooler, turbo, exhaust, driveshaft, rear differential, axles, fuel cell, and radiator.

the last two things i can really think to do is the window and hood molding. then afterwards all of this will be joined together for a single interior.joe file which will be a very drawn out process to texture. since i'm more of a modeler than a texture-er i always find myself saying "why am i fussing with this in a texture?? why not add more detail to the mesh??"

questions
1. i'm currently on a athlon64x2 @ 2.2ghz, 2gb ram, geforce 7900gt. built some ~18 months ago. i've been getting carried away with the polycount (to say the least) cause i'm at ~12500 polys. but my framerate (with the old engine) never drops from 85 (which is the refresh on the monitor). so i'm guessing this is all still good. am i right?

2. does anyone have a windows exe for a recent build that can run stable for like 5 minutes or so?? it would really help. the last release still defaults your keys everytime you restart, so when i do an ingame look around i have to remap everything first. also tempted to add more polygons to even out the lighting. but this might be useless with a specular map (don't know for sure, haven't worked with those much, but per-pixel specular sounds like it could do the trick).

3. i started throwing together a driver mesh. right now he looks like a goon, but is a driver mesh part of the plan? as in seperate from the car, can be plugged into any car??

now a requst:
not to sound like a broken mp3, but i think it would make the art generation easier if cars could be split into lots of different .joe files (have a model list in each .car file). doesn't have much impact on mesh making, but it would help texturing a whole bunch. 'cause then i could get the radiator looking nice, and map all it's uv coords. then move to the next thing. this is an extreme case (cause i've modeled tons of internals for when you see through the grills on front and top) but my actual interior looks weak, and will need lots of little tweaks to make it look nice. but as it stands now, when i add polygons to the seat then i'll have to redo the uv coords on my exhaust pipe (yay!!).

grill work

[thelusiv]

and after thinking about that for a little while i threw in some low-poly placeholder meshes for the engine, intercooler, turbo, exhaust, driveshaft, rear differential, axles, fuel cell, and radiator.

the last two things i can really think to do is the window and hood molding. then afterwards all of this will be joined together for a single interior.joe file which will be a very drawn out process to texture. since i'm more of a modeler than a texture-er i always find myself saying "why am i fussing with this in a texture?? why not add more detail to the mesh??"

I like the idea of modeling everything separately. I never am good enough at modeling to do that quickly, but I can do 2D stuff pretty fast, so for instance with the XS I just slapped a texture on to do things like hood, door handles etc.

1. i'm currently on a athlon64x2 @ 2.2ghz, 2gb ram, geforce 7900gt. built some ~18 months ago. i've been getting carried away with the polycount (to say the least) cause i'm at ~12500 polys. but my framerate (with the old engine) never drops from 85 (which is the refresh on the monitor). so i'm guessing this is all still good. am i right?

As long as you don't go super crazy with the detail, it should be fine. And by super crazy I mean, adding extra polys that aren't necessary like lots of people do when they model cars just for looks. I'd say that if the poly count doubled after you model the interior, it wouldn't be so bad. Not that you have to, if you can do it in less, of course.

2. does anyone have a windows exe for a recent build that can run stable for like 5 minutes or so?? it would really help. the last release still defaults your keys everytime you restart, so when i do an ingame look around i have to remap everything first. also tempted to add more polygons to even out the lighting. but this might be useless with a specular map (don't know for sure, haven't worked with those much, but per-pixel specular sounds like it could do the trick).

I for one don't have a Windows machine with a good enough video card to do serious VDrift testing on anymore. We've been looking for a Windows developer for a long time, and some people have shown interest, but nothing ever seems to pan out. There are compilation directions for Windows on our wiki, I bet you could get it working. If you can't, just stop by #vdrift or post your problems here...

3. i started throwing together a driver mesh. right now he looks like a goon, but is a driver mesh part of the plan? as in seperate from the car, can be plugged into any car??

If you want to develop this driver further that would be excellent. It would be nice to include a driver that works on every car. This would mean the driver would have to be made up of parts that can be positioned a little differently for each car. Later, we could work towards making the parts move with the motion of the car.

now a requst:
not to sound like a broken mp3, but i think it would make the art generation easier if cars could be split into lots of different .joe files (have a model list in each .car file). doesn't have much impact on mesh making, but it would help texturing a whole bunch. 'cause then i could get the radiator looking nice, and map all it's uv coords. then move to the next thing. this is an extreme case (cause i've modeled tons of internals for when you see through the grills on front and top) but my actual interior looks weak, and will need lots of little tweaks to make it look nice.

Tell you what, if you set up a car that can be used for testing creation of something like this, someone will probably do it. The tracks are already loaded in this manner. Go ahead and write a mock-up .car file that can be used as well. You could use the TC6 if you want.

but as it stands now, when i add polygons to the seat then i'll have to redo the uv coords on my exhaust pipe (yay!!).

I'm not quite sure why you have to do this. I haven't been using Blender a lot lately, but I seem to remember not having to re-map things constantly when I made changes to other stuff. How are you going about UV mapping?

The grills look quite nice! The car is progressing well. Have you worked on a .car at all yet?

[zimluura]

== UV-mapping ===================
when i do uv mapping i usually start off with the unwrap command. then re-position things a little bit.
blender does seem to compensate for edge and face cuts (shift+k), which could work. but its behavior is a little strange when you extrude. and i'm guessing it has some problems when you apply a mirror, then start tweaking one side differently from another. small stuff like a gas door (or the little thingy on the hood of the tc6) will likely cause a little bit of problem.


== .CAR File ===================
i've got a .car together. was pretty easy to start with the tc and then give it a little more kW. found a link on the forum here to a site with lots of wheel dynos, then used that and the stock crank kW/torque. approximating a 17% loss in the drivetrain. seems too fast though. when i times a 0-60 i got some 3.5 seconds. checked all the masses and they seemed good (little or no change since the 5th gen tc).

edit: maybe not enough power loss in the awd drivetrain. should i add add more mass to the wheels to compensate maybe??


== Component TC6 testbed ===================
ok, i'll work on splitting up the tc6 the way i'm thinking of...
...1.5 hours later

http://zimluura.googlepages.com/tc6_ext.zip

each has a single color, low res texture. should look sortof ok when it gets loaded. size is really reasonable, but that's 'cause there are no high res textures. only a few of them will need to be high res in the end.


== Polycount ===================
i don't foresee it putting on too many more faces. i think with another 1000 i would have the inside looking really really nice


== Driver model ===================
he looks quite a bit less goony now that he has a texture map, he's pretty simple and doesn't have fingers. fits in the tc6 but haven't tried him yet in the le. also not really sure if how tall he'd be in real life. just guestimating so far. thinking, future development-wise, at some point in the distant future a skeletal system for animated pit crews. but if there's a skeletal system there then that would be able to make the driver steer the car, look around, jump out of the car and run to safety as it explodes.


== VDrift windows build ===================
every so often i get an svn then kill some time trying to compile it in vc6. so far, how the includes are laid out is different enough to stop me in my tracks. i might give some time to that for a second to see if i can (it's the only ide i know, aside from turbo c++ 3 for dos). when that fails misserably i'll take a shot at following the instructions . but i guess this is what i should work on for a bit. if no-one else is on it.

[joevenzon]

maybe not enough power loss in the awd drivetrain

The AWD code is a bit of a hack so you might have to manually cut down on the power to compensate....

i don't foresee it putting on too many more faces. i think with another 1000 i would have the inside looking really really nice

Should be fine.

if there's a skeletal system there then that would be able to make the driver steer the car, look around, jump out of the car and run to safety as it explodes

Yeah, eventually I'd like to do this. The Cal3D library is an easy way to add skeletal animation.

every so often i get an svn then kill some time trying to compile it in vc6.

If I get a chance I'll try to compile a windows version for ya. What we really need is a proper windows release, though. :-)

[thelusiv]

The AWD code is a bit of a hack so you might have to manually cut down on the power to compensate....

http://code.google.com/p/vdrift/issues/detail?id=45

If I get a chance I'll try to compile a windows version for ya. What we really need is a proper windows release, though. :-)

I have a Windows test box now so I may get a Windows binary preview out soon.

[thelusiv]

when i do uv mapping i usually start off with the unwrap command. then re-position things a little bit.
blender does seem to compensate for edge and face cuts (shift+k), which could work. but its behavior is a little strange when you extrude. and i'm guessing it has some problems when you apply a mirror, then start tweaking one side differently from another. small stuff like a gas door (or the little thingy on the hood of the tc6) will likely cause a little bit of problem.

Hmm, I started off using unwrap, but I found it was easier to select a subset of the faces, and then use the "Project from View" method of UV mapping. That way you can go above the car, select the faces that mostly face the top, and then map them; look at the side of the car, select the faces that face in that direction, and map them; etc. Afterwards you can position the different sides in a way that is sort of like unwrapping them.

The trouble I had with unwrapping was that certain faces would be scaled differently than others, making textures get applied in an inconsistent manner. So for instance if I wanted a pinstripe that went down the side of the car, it would be thicker in certain parts and thinner in others. Thus I switched to projection because it kept things proportional to how they really were. Take a look at the XS texture (and the UV mapping in the .blend fine as well if you like) for an example of what I mean.

ok, i'll work on splitting up the tc6 the way i'm thinking of...
...1.5 hours later

http://zimluura.googlepages.com/tc6_ext.zip

each has a single color, low res texture. should look sortof ok when it gets loaded. size is really reasonable, but that's 'cause there are no high res textures. only a few of them will need to be high res in the end.

Could you submit an issue report about this?

On textures: should each individual model have its own texture? I think each object should have the texture set in the .car file. That way, some objects can share textures. Joe, what do you think?

every so often i get an svn then kill some time trying to compile it in vc6. so far, how the includes are laid out is different enough to stop me in my tracks. i might give some time to that for a second to see if i can (it's the only ide i know, aside from turbo c++ 3 for dos). when that fails misserably i'll take a shot at following the instructions . but i guess this is what i should work on for a bit. if no-one else is on it.

I wouldn't waste too much time on VC6. We have a working Windows build setup, it's a little tricky to get all the tools in place, but hopefully I'll improve the documentation a little soon.

[joevenzon]

Could you submit an issue report about this?

it's already in there as issue 3
http://code.google.com/p/vdrift/issues/detail?id=3

[joevenzon]

On textures: should each individual model have its own texture? I think each object should have the texture set in the .car file. That way, some objects can share textures. Joe, what do you think?

I agree, we'd want to define which texture to use per joe file in the car file somewhere.

[thelusiv]

I saw that issue after I posted that, and CC'd zimluura on it.

[zimluura]

Hmm, I started off using unwrap, but I found it was easier to select a subset of the faces, and then use the "Project from View" method of UV mapping. That way you can go above the car, select the faces that mostly face the top, and then map them; look at the side of the car, select the faces that face in that direction, and map them; etc. Afterwards you can position the different sides in a way that is sort of like unwrapping them.

The trouble I had with unwrapping was that certain faces would be scaled differently than others, making textures get applied in an inconsistent manner. So for instance if I wanted a pinstripe that went down the side of the car, it would be thicker in certain parts and thinner in others. Thus I switched to projection because it kept things proportional to how they really were. Take a look at the XS texture (and the UV mapping in the .blend fine as well if you like) for an example of what I mean.

that sounds like it will help allot. thanks.