Announcement

Collapse
No announcement yet.

[WALK-THROUGH] Fully Functional Vehicle (my first car).

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • replied
    hi,

    finally finished it (would have been yesterday but had a minor emergency on tuesday), anyway down to buisness. i have dropped the wheels slightly, unwrapped and layed out the UV (kind of a hatchet job though), created a basic texture (1024x1024 slightly bigger than i was going to use but i can always redo a smaller one at a later time), applied smoothing, and re-checked scale and model lacation in blender (scale is fine, model location is at 0,0,0).

    i also did a test import to UDK, here is the result

    as you can see the tris count is basicly the same (actually dropped by 20 tris) witch is good, and the buggy looks ok-ish, the only small problem i found was on the inside parts of the driveshafts (parts next to eginge) the normals were flipped witch is pretty easy to fix. i have not done a specular or normal map yet but will do them soon.

    so what do you think, do you think it's good enough?

    Leave a comment:


  • replied
    ok cool. i will drop the wheels a little bit and create a UV tomorrow. the car is already at a scale of 1UU = 1blender unit so thats all good, i have already centered the car but will do it again after i have dropped the wheels down, i will also delete the place holder driver but will create a spear copy of the blend file to do the driver at a later date.

    as for LOD's erm yeah

    Leave a comment:


  • replied
    that will do nicely Sir

    scale that looks good, once you've done the uvw's we can add a simple rig just to make sure the scale is what you need.

    the scale i use i 1uu = 1 blender

    ps we won't mention LOD's yet

    Leave a comment:


  • replied
    youve got to love a mk2 escort

    i just deleted the new front wheels and replaced them with slightly smaller versions of the new rear wheels witch has dropped the total cars tri count to 16599 witch means a total drop of 15990 tris. do you reckon that would be low enough, i could probably get rid of 400-500 more if i delete the spot lights and got rid of the wholes in the steering wheel if nesacery (looks kind of wierd without the spot lights though)

    on to scale, here is a picture of the grid in the scene with my car (i leave it at the default sizes, i have seen your tutorial where you explain how to change it but i usually just use the properties window to check size)

    as you can see the steering wheel is about twice the size of the grid (16 units) witch makes it about 32 units witch seems realistic, i did add a cube of 64x64x64 and it filled about 3/4 of the cocpit. here is a pic of the car with the properties window

    i am pretty sure i scaled it properly. now i have 2 issues with scale, the first is the size in UDK because a few people say 1UU=2cm and other people say 1UU=1cm and im not sure wich one to use. the second issue was the size of the driver witch is why i thought making a roughly to scale stick man driver was a good idea.

    so to sum up 3 questions
    considering i plan on having a pretty low poly driver and am going to smoot shade most of it do you think a tri cout of 16599 is low enough?
    did i apply scale properly? (im pretty sure i did as i did it the same way i scale static meshes and they look fine but better to sort it now rather then later)
    witch size scale do you think would be best?

    thanks

    Leave a comment:


  • replied
    good so far, as for the scale of things you need to be careful with blender as a lot of the time when you import it into udk it will be matchbox size,

    best thing to do is a quick size check, create a 64 unit box and export that, i show this problem or should i say how to set blender up scale/size wise in my videos.

    btw i learned to drive in a escort mk2 estate

    Leave a comment:


  • replied
    thanks for saying so

    ok, so i fineshed the wheels today so i thought i would show the progress i have made.

    lets start with the numbers.(the numbers i posted for the wheels yesterday wre for after i tried to lower the tri count)

    so the original front wheels had 1250 verts and 2520 tris
    the original rear wheels had 1500 verts and 3020 tris
    so the original car had a total tri count of 32589

    the new front wheels have 554 verts and 1104 tris
    the new rear wheels have 458 verts and 912 tris
    the new car has a total of 16983 tris

    thats a total drop of 15606 tris

    also i tried to scale the car but had some trouble, i will expalin. here is a picture to help show what i mean.

    i figured the best thing to do would be to scale it 1/1 (1 blender unit=1cm) but did not know how big the driver would need to be.
    so i decided to put a driver in the car, i measured my own bones then made a basic stick man using those measurments (as you can see in the pic, and just in case the head looks to big to you i thought it would be easier and save on the tri count to have a helmet instead of a face and measured a crash helmet i had lying around, it was 25 cm wide so made the head the same size). after i made the stick man i centered him in the car, then scaled the car up until it fit the driver because the drivers roughly the same size as me. the cars dimensions (in blender units) are 353 long, 232 wide, and 143 tall. just to put that into perspective a mk2 ford escort is 414cm long, 160cm wide, and 139cm tall witch means my car is pretty big but is as accurate as i can make it. there is a bonus though im going to keep a spare copy of the car with the stick man, that way when i try to make the driver i can model him in situe, witch means he will fit perfectly and i can get the various body parts (hands on the steering wheel for example) in the exact right place. then when i finish modelling him i can just delete the cars mesh and be left with a perfectly tailored driver.

    i would appreciate your thoughts on this stuff

    Leave a comment:


  • replied
    great going your doing a fine job

    Leave a comment:


  • replied
    Thanks SPOOF for the technical explanation and for using basic terminology so i could understand it easily(very well written).

    GEODAV i am still aiming for 15k tris, i did try to just lower the tri count for certain parts (wheels/suspension ect...) but i only managed to get rid af 5000 tris so decided to make new wheels and suspension. now i havnt finished the wheels yet but heres a small update on the suspension.

    old suspension
    per corner
    VERTS:1130
    TRIS:2244
    TOTAL TRIS(whole car)=8976

    new suspesion
    per corner
    VERTS:314
    TRIS:612
    TOTAL TRIS(whole car)=2448

    an instant drop of 6528 tris (witch isnt bad)

    Leave a comment:


  • replied
    thanks Spoof for the extreme technical side of poly/vertex counts, never even knew about that .

    i say stick to a 15k count as this is a happy medium for a game like UT3 with quite a lot of vehicles, now if its a racing game with just a few vehicles then of course you could bump the count up, but lets stick with 15k because then later you'll need the extras for the driver

    Leave a comment:


  • replied
    Btw, for the large door panels, bonnet and sides you can afford to use flat shading. Or add a few edge loops to flatten the panel and retain a smooth trim.

    Edit: I'll hand you back to Geodav 'cos I know nothing about cars

    Leave a comment:


  • replied
    Originally posted by smokey13 View Post
    not sure what you mean by (unless many of those elemnets are already seperated from the surface)

    i did not know UDK makes more verts than what i put in the model in blender (seems crazy to me that UDK would add so many verts, "RANDOM") any way thanks for the heads up

    Those pics give a good illustration of what happens. 76k versus 13k vertices

    The two questions are related, and not random at all. Smoothing groups, UV seams, separate materials and physically separate parts of the mesh all generate additional vertices due to the way vertex normals and uv coords are handled.

    Say you have two quads connected together at one edge, but bent at a 90 degree angle. If the quads are flat shaded then the four vertices of each face have their normals pointing 'straight out' from the face. If they're smooth shaded then the vertices that make up the shared edge have normals pointing at an angle that 'bends' between the two directions.

    When lighting is calculated for these faces it uses the vertex normals to create the shading, either completely flat for each face, or smoothly interpolating between the normals.

    Now, in Blender or any other modelling package, this setup appears to share the vertices (6 verts for two joined quads, because 2 verts are shared). However, because of the way these tools handle their data internally, the shared vertices of the flat shaded model actually require two sets of normals. This is hidden from the user to simplify the modelling environment.

    In a game engine this setup is too complicated and inefficient for realtime rendering on a GPU. It's much more efficient to keep things simple and have 4 vertices with their own normal instead of two verts with two sets of normals.

    This same principle applies to the UV coordinates, so a vertex that appears along the seam of separate UV islands will actually generate two vertices in the same position, but with different UV values.

    So, a cube with 8 vertices in Blender/Maya/Max will import into UDK as 24 vertices if all 6 faces use flat shading, or form separate UV islands, or have separate materials applied to each face. If they are part of a single smoothing group (all faces use smooth shading) then they should import as 8 vertices (but lighting will look very odd for such a low-poly mesh using smoothing).

    Leave a comment:


  • replied
    hi spoof, i was planning to smooth shade most of it (just leave a few flat shaded surfaces like the edge of the wheels). i imported as a stitac mesh as you suggested and here's a pic of the results

    the one on the left is all flat shaded, on the right smooth.

    as this is my first try im not going for proper realistic, just learning how to do it. im not sure how close you wil see it because im going to release this first car as a free asset when it's finished (with scripts and everything) so anyone can just drop it in a level and start driving, although i did have a really good idea for a racing game and if i can figure all this out i will use what i learn building this car (and in racing games you get very close to the other cars so need to learn to do suspension). i am also going to try and stick to GEODAV's budget of 15k tris.

    as for current polycount
    the springs have 1744 faces per spring ,they have an octogan profile and 12 sections per coil. i am currently modeling there replacements, i amgoing to have a square profile at a 45 degree angle and only 6 sections per coil, im also making them thicker so i wont need as many coils (this should cut the number of faces by roughly 65%).

    and the rear wheels have the highest polycount with 2380 faces per wheel and the front wheels have 1330 faces but im making a whole new set of wheels with a simpler tread and a much simpler rim design witch should cut the polycount even further.

    not sure what you mean by (unless many of those elemnets are already seperated from the surface)

    i did not know UDK makes more verts than what i put in the model in blender (seems crazy to me that UDK would add so many verts, "RANDOM") any way thanks for the heads up

    Leave a comment:


  • replied
    Also, with this kind of hard surface modelling, your vertex count is likely to increase quite a bit on import (unless many of those elements are already separated from the surface). I would definitely try importing it straight away just to get an idea of the actual numbers.

    Edit: just looked at the pics again. You need to at least set the tyres to smooth, otherwise every single quad will generate 4 vertices on import.

    Leave a comment:


  • replied
    I'm guessing most of your polys are in the springs. You could minimise that by making the spring use a simple triangular profile, smoothing should fool the eye into seeing a tube.

    However, a more optimal trick is to make the whole spring section a simple tube and fake it with a double sided alpha texture. You'll still get the flexing animation when the tube bends

    It all depends how realistic you want it, how close the viewer will get, how big your overall poly budget is.

    Leave a comment:


  • replied
    ok, lowering the polycount of the wheels and front tyres shuold be easy enough(the front tyres actualy have 100 faces around the edges alone (was 50 before i triagulated)times by 16 edges of each front tyre witch =1600 tris just for the part of the tyre that touches the road), the rear tyre will be more difficult due to the tread. also i was thinking of using flat shading on the frame witch means i can get rid of a few tris there, and i can also ditch a few of the asthetic body mounts witch will help. as for the bones you explain about blender's alignment issues very well in your tutorial so i think i will be able to align the bones to the correct orientation as long as i follow your tutorial taking each step nice and slow. as for compressing the spring i too believe we can get it working.

    Leave a comment:

Working...
X