#SKIN – body with legs?
39 messages in this thread
Gary,
I'm a little confused about the procedure to create branched forms such as legs
using SKIN. Is this possible or can SKIN only work with topological cylinders
and not forks?
Thanks,
Terry Gilbert
Forcade & Associates
Although skin.axp doesn't skin forking branches, there are certainly many
things you can do with texture maps, etc that'll mitigate the forking spot.
– G
Gary:
/__________From your message___________\\
that'll mitigate the forking spot. __________________________________________
C'mon Gary watch your language, this is a family forum. <GGGGG>
Bob
Will the R4 skin routines be able to cover a situation such as this? I have
read all the stuff posted on-line, but I am still a little fuzzy on some of it.
Just how different from skin.axp on disk #6 is it going to be? What can it do
differently? I assume since you made the announcement you can talk about these
types of things.
R4 comes with a Bezier patch modeler, which is plenty different than a skinning
routine. For more info about r4, check out r4feat.text in the 3DS DL.
– G
I've read all the stuff in the libraries abour R4. I wanted some new
information. Can you give a quick example of how the new bezier patch
modeler could be used to create something? Or anyone else who can explain
this, actually. I think I know how it is supposed to work, but I am fuzzy,
and need more enlightenment.
I think I'll leave that description to either Jonas, Nik, Brian, Phil or
one of the other Autodesk folks.
– G
If you want someone to give you the low-down on bezier-spline patches in
general, I'd recommend asking M. G. BATCHELOR 71532,1214 — he seems to
know quite a bit about stuff like that (you'll remember he wrote the tips
section at the end of the SKIN readme).
Dave
<< If you want someone to give you the low-down on bezier-spline patches in
general, I'd recommend asking M. G. BATCHELOR >>
Gee, thanks David…but I didn't go to SIGGRAPH & haven't seen the Plug-In.
I really shouldn't say anything then, ….. but simply referring to the PR:
1) Sounds like you begin with a planar or cylindrical Bezier patch mesh.
2) Controls are provided to manipulate CV's individually or in sets,
including the ability to locally deform the patches. Good local control of
deformations would allow more automated/powerful manipulations of CV's than
3DS currently provides in the Editor with vertices. For example, set 2
points to anchor the patch at 2 sides, and set another to pull out from the
center – stuff like that….may use Bezier handles to accomplish this with
CV sets, I don't know.
3) Sounds like you can snap patch CV's to Polymesh vertices….if so, you
could "drape" a mesh to conform to an existing object (manually), or sort
of "shrink wrap" a patch around an object.
4) Sounds like the patches are "intelligent", in that their edge resolution
(after tessellation) automatically (?) varies as needed with distance from
camera in order to avoid stepped edges.
5) Sounds like the patches can be blended together, that is, "sewn"
together at the edges while maintaining tangency & continuity. Typically, a
complex surface would be constructed out of many patches.
6) Sounds like they can be mapped parametrically….but I don't know what
the text means by "automatically OR parametrically".
AGAIN, I'M ONLY GUESSING AT THE SPECIFICS HERE.
BILL
Not bad for someone who's never seen it! <g>
– G
>> sort of "shrink wrap" a patch around an object. <<
Clothes, bedspreads, etc….
Bill:
I _said_ about spline patches in GENERAL! Am I supposed to believe you're not
beta-ing r4 (at least the patches), though?<g> I know you'd be under an NDA if
you were…. Me, I'm just under a rock… or the gun… or behind the
8-ball….
Dave
<< I _said_ about spline patches in GENERAL! Am I supposed to believe you're
not beta-ing r4 (at least the patches), though?<g> >>
No, I'm not testing anything for R4 – honestly. If I was, you asked, and I
couldn't comment – I'd just say "I can't comment". But in this case, I can
honestly say I don't know as much as most of you here – 'cause I didn't go to
SIGGRAPH, and I'm really not testing any R4 stuff.
If there's something specific anyone would like to know about Patch Modeling,
I'll try to answer it. But in general:
The Bezier patch is just a net or matrix using Bezier curves as boundary curves
to define the patch. The advantages of using them in model construction can
best be described with an illustration. Take a 3X3 polygon grid…if you wanted
to model this into a flying carpet-looking thing, first you'd need a much
higher resolution grid mesh to begin with – and thus a lot more vertices to
deal with. Then, to approximate the gracefully sweeping contours, you'd have to
perform tedious multiple vertex selections, and apply rotations to form the
graceful contours…..a pain.
If you displaced a single vertex in a 3×3 polygon mesh, you'd get a sharp
spike, and likely get shading problems. With the patch, displace a CV and you
get a graceful contour, which would take a much higher-res polymesh and a lot
of work, as I said. You could model your flying carpet with a few
selections/displacements. Also, the patch modeler has to deal with maintaining
tangency & continuity at the intersection of patches "sewn together" –
initially & dynamically, as the composite is deformed. Again, this would be
very difficult & tedious if attempted with several polymeshes manually.
Finally, you (or the program) has the ability to variably subdivide the mesh at
render time, and this can be used to perform the adaptive tessellation based on
proximity to the camera, as implied in the PR. Also, interpolation of the
normals generally should provided a smoother shading than working with
polymeshes, or the curved surface could even be rendered as a surface, rather
than a bunch of triangles.
So with a decent system, you can see the curves as you model and can control
the UV density for visulaization purposes, are only working with a minimum
number of CV's at a time which produce similar results to many polygon vertices
& manipulations, and since you're working with curves, there should be more
powerful deformation tools to help quickly "mold" the patches as desired. The
system takes care of the normals, patch subdivision, and maintaining surface
continuity – or you can override the system. Finally, texture mapping can be
automatic, causing the map to conform to the contours of the patch mesh.
I personally like working freeform with NURBS curves/surfaces, but a good patch
modeler can be a very powerful addition in the right hands, and should be a
most welcome addition to 3DS.
BILL
Bill,
I've been following these SKIN and patch threads for a couple days, and am
checking out the samples you uploaded. I hope to use SKIN to drive
captured-motion characters for game development soon. The woman.zip looks
really great!
You've done a great job of explaining Bspline patches. Could you explain what
NURBS is and how it is different from Bspline patch modeling?
Thanks.
Eric Chadwick @ Mechadeus
Hey, thanks for the compliments Eric.
<< Could you explain what NURBS is and how it is different from Bspline patch
modeling? >>
Whew ! That's tough, without explaining spline types, etc…..also I'm no
mathematician, but here goes from a user's point of view. First of all, there's
various types of curves distinguished by their individual characteristics and
these various types of curves can define patches named after them. You can
model freeform using these control curves along with various functions to
assist you, and define surfaces, patches, etc.. That's a freeform modeling
approach. With a patch modeling approach, you start out with one or more types
of pre-defined patches (sorta' like grids made of spline curves & their CV's as
opposed to polygon vertices & edges), and then use the CV
manipulation/deformation tools to form the patch into some object.
NURBS curves (and I suppose, Bezier) are types of B-Spline. B-Splines can be
classified as rational, uniform & non-uniform. A NURBS curve is a Non-Uniform
Rational B-Spline, and is generally the most sophisticated curve to work with.
The "non-uniform" term comes from the fact that you can construct curves with
non-uniform knot spacing. The knotpoints I suppose are the points at which the
individual curve segments are joined into the composite curve you define by
defining the control lattice with control vertices. The rational term comes
from the way they are defined mathematically – as a RATIO of two polynomials.
They are "better" than Uniform B-Splines because: 1) They interpolate better
through control points… 2) Certain types of curves such as true circles can
be constructed… 3) They incorporate "weighting" as an additional local shape
controlling feature, allowing more subtle & precise curve construction.
Bet you're sorry you asked me now, huh ? <G> That's absolutely as succinct as I
could put it, sorry.
Regards,
BILL
>>I personally like working freeform with NURBS curves/surfaces,
Out of curiosity, have you tried the demo 3ds NURBS modeller in the
libraries here? I haven't grabbed it yet, but I'd be curious to know how useful
it would be.
David
Hi David,
Yes, I did try it out for a couple of days. With all due respect to the
authors, I wasn't crazy about it because as I've said before – I *prefer*
free-form spline modeling from control curves, and it wasn't nearly as
full-featured as any that I've worked with before. On paper anyway (I haven't
seen it), the R4 patch modeler should be much more useful, even though it's
Bezier vs NURBS. (IMO)
BILL
Tim,
<< how the new bezier patch modeler could be used to create something? Or
anyone else who can explain this, actually. I think I know how it is supposed
to work, but I am fuzzy, and need more enlightenment. >>
Bill's description is very close. Dealers will be receiving Beta copies late in
September and you'll be able to see it then.
when can we expect an official release for the rest of us? I have heard end of
Sept, end of Oct, and end of November. Can I shoot for October?
Tim,
<< I have heard end of Sept, end of Oct, and end of November. Can I shoot for
October? >>
The recent public notices have stated Oct.
>>
Although skin.axp doesn't skin forking branches, there are certainly many
things you can do with texture maps, etc that'll mitigate the forking spot.
^^^^^^^^
<<
Gary, such language!
Terry Gilbert
Forking & Associates
Boy oh boy… I'm in trouble now.
– G
Gary / ALL,
<< I'm a little confused about the procedure to create branched forms such
as legs using SKIN. Is this possible or can SKIN only work with
topological cylinders and not forks? — Although skin.axp doesn't skin
forking branches, there are certainly many things you can do with texture
maps, etc that'll mitigate the forking spot. >>
PMJI, but I've fooled around with this sort of thing a lot since Skin beta,
and actually you can……sorta'. As discussed in a recent thread, this
requires some misdirection though <g>. Of course you can not create nor
sweep skin along a true branching path which is concatenated to a single
vertex at the branch point, but you can fake it pretty nicely using a
technique similar to back-tracking paths and Skin parametric capping
techniques. You may or will end up with a single "seam", but you can
displace it & somewhat hide it to a less obvious location than the *actual*
joints.
There are a number of ways to accomplish this depending on the need of the
subject model, but I'm uploading an EXTREMELY simple example as
"Branch.zip". This contains 2 small flics & 2 project files. The subject is
merely a cylindical stick-man body & legs (an upside-down 'Y').
"Branch.prj" shows the simple construction using quad CS's skinned by
prefix, and the flic just rotates it 360 on Y to show the lack of seams
which would normally show all around the "thigh/pelvis" intersection.
Select by name each CS and analyze this technique, as it won't be obvious
otherwise. The seams are reduced to a single seam at the "crotch" in this
example.
"Branch2.prj" is the same 'Y' stickman body & legs in a walk-cycle loop,
using a skeletonal heirarchy for control of the CS's rotations and
"seamless" joints. The flic shows that (at this scale anyway), NO seams are
perceivable when in motion – part of the misdirection. With variations on
this technique, motion/framing, mapping, etc. – you can construct a very
detailed (apparently) seamless character. I've been working on/with a VERY
detailed Skinned female body with modeled muscleature (no bump maps), and
all natural "seamless" joints for some time now. (except for the hands,
which are very difficult)
Regards,
BILL
Wow! Amazing (as usual), Bill. I had a feeling that it was possible to
"fake" it like that, but nobody had done it yet (until you did, of course)
and I wasn't sure.
I'll go grab the file now and Jack and I'll tear it apart and figure it out
this afternoon.
I can't wait to see your detailed skinned female body… (how detailed?
:-}
– G
<< I had a feeling that it was possible to "fake" it like that, but nobody
had done it yet >>
Don't expect too much, it's just a little wishbone lookin' thing….but
illustrates the concept. Basically the concept is in laying out CS's along
your "skeletonal path" in the Editor, then back-tracking down one of the
branches to be in position to continue down the other branch. You can use
same-sized/positioned CS's, scale down & create a hidden inside surface
'til you find a suitable point to come out, rotate CS's as necessary hidden
inside the visible surface, etc…Motion, lighting, mapping, framing, etc.
is VERY important here to hide the inevitable seams, but you can handle a
lot of things with this technique. Mapping can get very difficult.
<< I can't wait to see your detailed skinned female body… (how detailed?
:-} >>
Very….but PG-rated <g>. This is for the first *job* I've taken in a long
time, and the first all-3DS construction I've done commercially. I'm very
pleased with it so far. If *my* body would just hold up so well, to allow
me to finish it on time <g>. Maybe I can u/l renderings & wireframe, but
there's no head yet – can't u/l the mesh though, at least at this juncture.
It's designed to be IK driven (I guess). BTW, I forgot to loop that
wishbone walk-cycle example flic….sorry about that.
BILL
A commercial project… I can just imagine how much detail you're going to put
into THAT.
I'm looking forward to seeing the pics.
– G
Wow! Very very nice model. Whoever is paying you to do it a lucky customer.
(great proof of skin.axp as well)
I'm looking forward to seeing her when she gets her head, too! <g>
– G
Gary / ALL,
Hope I don't get into hot water with this….but anyway, I've uploaded 2 views
of the still-headless female Skinned Skeletonal-driven mesh and 1 wireframe
screen dump. The file is called "Woman.zip", and is small enough (< 70k) that
anyone interested can d/l it. This is from last week's version which is not
entirely seamless, but it's hard to tell with these renderings. Like the cowboy
head Skinned model, this is a very efficient mesh. I believe this version was
less than 3500 vertices total, and about 1800 without hands & feet (considering
that half the vert's in a Skin CS are ignored, right ?) This model uses no bump
mapping, and uses the complementry ambient/specular technique you described
recently to help bring out the definition. If too bright/washed out, turn gamma
corr. off or reduce.
WARNING: These images are PG-rated in that *_certain areas_* of the body are
not completely anatomically correct <g>, but she still doesn't have any clothes
on.
BILL
I'll see about getting them merged into the DL's asap. Thanks for uploading
them!
– G
Thanks for all the nice comments Gary….if I can just maintain long enough to
finish this thing. BTW, let me publicly retract my statement to you regarding
difficulty with your metal shader/refmapping. It took an evening of
experimenting to get what I was looking for, but now that I've acclimated to
your system a little more, I've gotten great results. More than once you &
others have praised Renderman here, but I tell you, your renderer is one of the
best I've ever seen, has a similar RM "look" & then some, and is considerably
faster. Those guys deserve a lot of credit for pioneering this stuff, but 3DS
just keeps impressing me – the more I get to work with it. BTW, the tape was
very nice. I don't think most folks would know (believe) it was produced with
PC s/w. Signal quality held up well, too.
BILL
Looks like I lost your support on inproving the render for the next rev. Is any
one looking at Alias or Wavefronts render for a side by side comparison? I know
my people are. regarding the nexus jpg uploaded to asoft I found the glass to
be PASTY and DULL and the gold to be less than perfect. going to D2 and
rendering feilds will not help me at game development sorry.
Bill I hope you are feeling better soon. Brandon
No, I didn't say the renderer couldn't be improved <g>. But having always felt
that 3DS looked kinda' pasty myself, I just recently discovered the proper
combinations to overcome that look pretty nicely with this particular program.
It did not come to me as intuitively as with some other programs. With the
combination of the right Medit surface parameters, the right map at the right
contrast, the right combo. of lighting, and especially the "right" gamma
correction…..I was able to achieve very nice polished metallic looks….it
kinda' surprised me when I hit on all the right elements. Personally, for
rich/warm looks like many lusterous metallic treatments are, I run gamma
correction lower that 1.8….that alone can make a big difference. And viewing
a NTSC framebuffer output as opposed to SVGA makes a slight difference in those
types of looks also. The mesh and it's mapping assignments are important too,
to bring out the look. The ambient material setting is very critical, moreso
than some other programs – and there's a narrow window of useable looks wrt the
highlight size/intensity (for lusterous metallic looks). As far as transparency
goes, most every scanline renderer I've seen looks pasty wrt ray tracers. Some
RT's look pasty.
Regards,
BILL
I too can get that look [rt] buy working at it, but I think it would be to the
Yost's group advantage to include some sort of selective raytracer my stomach
sank when r4 did not include a rendering upgrade. Brandon
Bill, coming from you (one of the most critical folks around here) that's a
REAL compliment.
As you can see from the demo reel, you're not the only one who's figured that
out. <g>
– G
PS: I'm glad you noticed the signal quality on that vhs tape. I can't believe
how many hoops I jump through to maintain quality. Working with vhs is brutal,
but not impossible.
<< ….that's a REAL compliment. >>
Y'all deserve it.
BILL
Gary,
I think the demo tape is THE BEST QUALITY IMAGE I have ever seen on VHS. Even
on the dinky cheapo TV/VCR combo unit in my kitchen the images are amazingly
clear and vivid.
Watching on a studio monitor from the original master must be dazzling!
Alec Jason
You ought to see it on a Sony digital monitor coming off of Digital Betacam. I
had to bring extra shorts to the editing sessions.
– G
Watching it at the Autodesk party on the huge monitors they had was stunning to
put it mildly.
ME
<< If too bright/washed out, turn gamma corr. off or reduce. >>
Geez, what I meant to say was Input File Gamma should be set to around 2.2 if
your SVGA display is set at 1.8.
BILL