 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
scott wrote:
>>> Like I said,
>>> build a rough mesh by hand with maybe 100 or so polygons
>>
>> Uh... wouldn't that still be absurdly time-consuming?
>
> No, not with all the tools in 3D packages specifically aimed at drawing
> meshes from scratch. Like face/edge extrude, multiple selection and
> operation on vertices/edges/faces, mirroring ummm countless others I'm
> sure.
That's like saying "yeah, the GIMP has functions to draw flat colours,
blur edges, and copy sections of an image". Well, sure, it *does*... but
how do you do useful stuff with that? (Clearly it's possible - but I've
never figured out how. It's highly non-obvious how these features "help"
in any way.)
> Here's a screenshot of a car I'm working on by drawing the mesh
> from scratch. You can see I have the front wing shown in wireframe so
> you can see the vertices, it really doesn't take very long to get them
> lined up nicely with a reference drawing. After that the modeller will
> smooth it out for me by replacing each face with perhaps 4 or 16 faces
> in the correct places for a nice smooth finish.
Jesus, how did you get so many points all spaced out so perfectly like
that?? o_O It must have taken you *hours*!
Also... just placing them in 2D is no good. You have to somehow position
them in 3D space - which (in every package I've seen) is absurdly hard.
> Actually, using this method with NURBS surfaces is very similar to
> adjusting the vertices by hand of a rough mesh and watching how the
> smoothed mesh changes. The number of control points required in NURBS
> is quite similar to the number of vertices needed in a rough mesh to get
> similar control of the final surface.
The difference being that NURBS produce a well-defined mathematical
surface which can be drawn at arbitrary resolution and can contain sharp
corners, whereas a polygon mesh presents only a very crude approximation
to such a surface (unless you generate terabytes of mesh data).
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>> Here's a screenshot of a car I'm working on by drawing the mesh from
>> scratch.
>
> I wish I had your talent with car modeling, you've posted some great stuff
> in the past.
Thanks!
I learned by following some (blender?) tutorials about drawing cars from
blueprints. You can get the blueprints from a lot of different websites,
then you set up the 2D views in Blender so you can essentially just trace
the lines in the blueprint with vertices and edges, making sure they line up
in all views.
After a while you get faster at doing this and learn how to use the
different tools in Blender. It took me a while to figure out how to use the
mirror tool properly, so you only need to draw half the car and it correctly
merges the central join with normals nicely as you go along.
I just need to work out how to properly constrain certain vertices and edges
during the automatic smoothing stage, sometimes it rounds off corners too
much that I want to stay sharp.
Blender is a PITA to learn sometimes, the GUI is just not in line with every
other application for Windows, but if you persevere with it you can get some
stuff done in it! The really weird file open and save dialog boxes annoy
me, the seemingly random way the different "panes" in the toolbars get half
hidden and put on top of each other, the weird way that program settings are
saved with each model and not with the program itself, I could go on and
on... But it's free!
Anyway this model was meant for a game, so I want to keep the low-poly
version looking as realistic as possible, as a side process I will try to
create a higher-mesh version for export to POV (of course) :-)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Jesus, how did you get so many points all spaced out so perfectly like
> that?? o_O It must have taken you *hours*!
I started with a n x n grid of vertices in a plane... ( a blender primitive)
> Also... just placing them in 2D is no good. You have to somehow position
> them in 3D space - which (in every package I've seen) is absurdly hard.
...and then in blender you can split your view into 4 so you have left, top,
front and 3D all visible at the same time with the appropiate 2D image you
are copying from in each background. It's then remarkably simple to just
drag each vertex to where it should be.
> The difference being that NURBS produce a well-defined mathematical
> surface which can be drawn at arbitrary resolution and can contain sharp
> corners, whereas a polygon mesh presents only a very crude approximation
> to such a surface (unless you generate terabytes of mesh data).
You do realise that NURBS objects are converted to meshes before being
rendered, even in POV? Surprisingly with smoothed normals you don't need
that much mesh data to get a *really* smooth look, 100k triangles for a car
and you'll be very hard pushed to see any mesh artifacts.
http://www.greatwallofgaming.com/wp-content/uploads/2007/07/gt5-prologue-5.jpg
That is rendered in realtime on the PS3, see any triangles in the car
models? (I think I remember reading somewhere the models are about 100k
triangles per car). I can just about make out the mesh around the front
wheel arch, but if they just applied 1 extra level of subdivision for a
rendered shot that would disappear.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
scott wrote:
>> Jesus, how did you get so many points all spaced out so perfectly like
>> that?? o_O It must have taken you *hours*!
>
> I started with a n x n grid of vertices in a plane... ( a blender
> primitive)
I find that whenever I do this, the points that haven't moved yet look
all perfect and regular, and the points I've moved by hand look all
wonky and out of place. A bit like when you try to draw a straight line
freehand next to some lines drawn with a ruler...
>> Also... just placing them in 2D is no good. You have to somehow
>> position them in 3D space - which (in every package I've seen) is
>> absurdly hard.
>
> ...and then in blender you can split your view into 4 so you have left,
> top, front and 3D all visible at the same time with the appropiate 2D
> image you are copying from in each background. It's then remarkably
> simple to just drag each vertex to where it should be.
Except that in the other plane, all the points lay exactly on top of
each other, so you've got to select the correct point in one view, and
then drag it in another view, then select the next point in the first
view, and drag it in the second view, and.......
>> The difference being that NURBS produce a well-defined mathematical
>> surface which can be drawn at arbitrary resolution and can contain
>> sharp corners, whereas a polygon mesh presents only a very crude
>> approximation to such a surface (unless you generate terabytes of mesh
>> data).
>
> You do realise that NURBS objects are converted to meshes before being
> rendered, even in POV?
Actually, I didn't realise POV could do NURBS at all! o_O
> Surprisingly with smoothed normals you don't
> need that much mesh data to get a *really* smooth look
...if you don't mind not having sharp edges...
> 100k triangles
> for a car and you'll be very hard pushed to see any mesh artifacts.
>
> http://www.greatwallofgaming.com/wp-content/uploads/2007/07/gt5-prologue-5.jpg
>
> That is rendered in realtime on the PS3, see any triangles in the car
> models? (I think I remember reading somewhere the models are about 100k
> triangles per car). I can just about make out the mesh around the front
> wheel arch, but if they just applied 1 extra level of subdivision for a
> rendered shot that would disappear.
Heh. Well the *track* certainly has plenty of polygons in evidence! :-P
I find it astonishing that there is any hardware on earth that can
actually render several hundred thousand triangles in less than a
minute, never mind a second. And yes, with the lashings and lashings of
trickery applied, you "almost" can't see them in this shot.
Let's see, what have we got? Well, first of all there's the shadow of
the wing mirror. Both wheel rims have jaggid edges. The front wheel arch
and the headlights less so. (Presumably when it's flying round the track
you don't get time to notice these things.)
Certainly it beats the stuff I have to watch in CSS all day! You don't
have to look very hard to find polygons *everywhere* in just about every
computer game I've ever seen. But then, it has to be realtime...
I still find it completely incomprehensible that somebody sat down and
spent months placing 100k individual triangles by hand though...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Invisible escreveu:
> scott wrote:
>> I started with a n x n grid of vertices in a plane... ( a blender
>> primitive)
>
> I find that whenever I do this, the points that haven't moved yet look
> all perfect and regular, and the points I've moved by hand look all
> wonky and out of place.
You should take a look at mirroring and proportional editing tools.
Also snapping to the background grid or fixing the editing to a plane or
axis helps.
> Except that in the other plane, all the points lay exactly on top of
> each other, so you've got to select the correct point in one view, and
> then drag it in another view, then select the next point in the first
> view, and drag it in the second view, and.......
It's not as complicated as you're saying. But you should know you're
way through 3D navigation well, because to model things you indeed most
likely will be rotating about it a lot to have better perspectives about
how the modelling is proceeding.
>>> The difference being that NURBS produce a well-defined mathematical
>>> surface which can be drawn at arbitrary resolution and can contain
>>> sharp corners, whereas a polygon mesh presents only a very crude
>>> approximation to such a surface (unless you generate terabytes of
>>> mesh data).
>>
>> You do realise that NURBS objects are converted to meshes before being
>> rendered, even in POV?
>
> Actually, I didn't realise POV could do NURBS at all! o_O
It doesn't. Anyway, subdivision surfaces produce roughly as good and
smooth surfaces as NURBS out of less detailed polygonal shapes.
>> Surprisingly with smoothed normals you don't need that much mesh data
>> to get a *really* smooth look
>
> ...if you don't mind not having sharp edges...
You don't see them with dense subdivided polygons.
> I find it astonishing that there is any hardware on earth that can
> actually render several hundred thousand triangles in less than a
> minute, never mind a second. And yes, with the lashings and lashings of
> trickery applied, you "almost" can't see them in this shot.
Oh, come on! Your 3D card in the computer already does that about as
well as a PS3 by now.
> I still find it completely incomprehensible that somebody sat down and
> spent months placing 100k individual triangles by hand though...
Nobody does that, not even in the early 3D days in the 70s where they
got models data out of rough 3D scanning processes.
Today, we use subdivision surfaces out of rougher polygonal models
extruded (you can extrude several faces at once) and bend and shape them
with various symetrical tools, like mirroring, and proportional editing
tools. You can also use sculpting tools and virtually go shaping a
dense mesh as if you were stroking it with a brush:
http://blenderartists.org/forum/showthread.php?t=143168
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Fri, 23 Jan 2009 10:10:27 +0000, Invisible wrote:
>>> Other people can pick up a pencil, draw a few brisk strokes on a pad
>>> of paper, and produce something the like of which I couldn't copy if I
>>> had several hours to work on it.
>>
>> Very few people have the raw talent to just pick up a new medium and be
>> good at it, for most of us we have to *practise* until we get good.
>
> I spent 6 months on taught drawing classes. You know what? I *still*
> suck at drawing.
And I sucked at playing the violin after 6 months. After 15 years, I
wasn't bad, but I haven't played in a while, so now I suck at it again.
You can't expect to become Da Vinci after taking classes for only 6
months. Some people study for years (or their entire lives).
Jim
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
scott wrote:
> simple to just drag each vertex to where it should be.
It's still very difficult and time-consuming. On the other hand, how long
does it take to write a decent program? That's difficult and time-consuming
too.
I'm going to have to learn Blender, I think. H-A:M is just too flakey to be
valuable as a commercial product.
> That is rendered in realtime on the PS3, see any triangles in the car
> models?
That's part of what the whole Phong shading stuff is about, yes, if I have
the right term? Where you don't shade the entire polygon the same, but
rather weigh each pixel's color based on the colors at the corners of the
triangles and the distance of the pixel from the corners?
--
Darren New, San Diego CA, USA (PST)
"Ouch ouch ouch!"
"What's wrong? Noodles too hot?"
"No, I have Chopstick Tunnel Syndrome."
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Jim Henderson wrote:
> On Fri, 23 Jan 2009 10:10:27 +0000, Invisible wrote:
>
>>>> Other people can pick up a pencil, draw a few brisk strokes on a pad
>>>> of paper, and produce something the like of which I couldn't copy if I
>>>> had several hours to work on it.
>>> Very few people have the raw talent to just pick up a new medium and be
>>> good at it, for most of us we have to *practise* until we get good.
>> I spent 6 months on taught drawing classes. You know what? I *still*
>> suck at drawing.
>
> And I sucked at playing the violin after 6 months. After 15 years, I
> wasn't bad, but I haven't played in a while, so now I suck at it again.
>
> You can't expect to become Da Vinci after taking classes for only 6
> months. Some people study for years (or their entire lives).
No - but you'd expect *some* improvement, surely?
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On Fri, 23 Jan 2009 18:39:20 +0000, Orchid XP v8 wrote:
> Jim Henderson wrote:
>> On Fri, 23 Jan 2009 10:10:27 +0000, Invisible wrote:
>>
>>>>> Other people can pick up a pencil, draw a few brisk strokes on a pad
>>>>> of paper, and produce something the like of which I couldn't copy if
>>>>> I had several hours to work on it.
>>>> Very few people have the raw talent to just pick up a new medium and
>>>> be good at it, for most of us we have to *practise* until we get
>>>> good.
>>> I spent 6 months on taught drawing classes. You know what? I *still*
>>> suck at drawing.
>>
>> And I sucked at playing the violin after 6 months. After 15 years, I
>> wasn't bad, but I haven't played in a while, so now I suck at it again.
>>
>> You can't expect to become Da Vinci after taking classes for only 6
>> months. Some people study for years (or their entire lives).
>
> No - but you'd expect *some* improvement, surely?
It depends on a lot of factors, not all of them in the student's
control. If the teacher is ineffective (either overall or for an
individual student's learning style), then there may be little or no
improvement, for example.
Jim
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
nemesis wrote:
> It's not as complicated as you're saying. But you should know you're
> way through 3D navigation well
I always thought it would be way better to have a 3D input tool, like Spex
gloves or something (or like the bits in the Minority Report movie).
--
Darren New, San Diego CA, USA (PST)
"Ouch ouch ouch!"
"What's wrong? Noodles too hot?"
"No, I have Chopstick Tunnel Syndrome."
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
