Hi.
Over the last few weeks I've been working on some cloth simulation
using Hugo Elias' method:
http://freespace.virgin.net/hugo.elias/
After seeing the cloth related discussion in pbi, I thought I'd post a
sample of where I've got up to.
The actual simulation was done in a C++ program (I did try a
pov-script one but have now left that for the 5000%+ speedup that
compiled code gave me...).  The cloth here is a 50x50 grid, with each
square split into two smooth_triangles based on maximum surface area.
On the left near the end of the animation you can see my current
problem - the jagged edges to the shadows caused by it still being
made of triangles under those nice curves.  I've been wondering
whether to do it with bicubic_patches instead, but that'll kill my
nice fast render times (7-8s per 400x300 antialiased frame on a PIII
600).
There is no randomness in this at all currently, hence the
super-symmetricality of it (it's not quite, actually, because it's not
centred exactly on the sphere), and the fact that it looks like it's
falling through a vacuum.
Well, that's probably enough rambling...comments welcome.

-- Michael Langton (first post here, BTW)

Post a reply to this message

Attachments:
Download 'ml_cloth1.m1v.mpg' (320 KB)

WONDERFUL, I am quite impress.
But I don't understand something: You modify the shape of
your mesh inside povray OR with another soft and you use 
povray to render this serie of mesh?

Fabian.


> 
> Hi.
> Over the last few weeks I've been working on some cloth simulation
> using Hugo Elias' method:
> http://freespace.virgin.net/hugo.elias/
> After seeing the cloth related discussion in pbi, I thought I'd post a
> sample of where I've got up to.
> The actual simulation was done in a C++ program (I did try a
> pov-script one but have now left that for the 5000%+ speedup that
> compiled code gave me...).  The cloth here is a 50x50 grid, with each
> square split into two smooth_triangles based on maximum surface area.
> On the left near the end of the animation you can see my current
> problem - the jagged edges to the shadows caused by it still being
> made of triangles under those nice curves.  I've been wondering
> whether to do it with bicubic_patches instead, but that'll kill my
> nice fast render times (7-8s per 400x300 antialiased frame on a PIII
> 600).
> There is no randomness in this at all currently, hence the
> super-symmetricality of it (it's not quite, actually, because it's not
> centred exactly on the sphere), and the fact that it looks like it's
> falling through a vacuum.
> Well, that's probably enough rambling...comments welcome.
> 
> -- Michael Langton (first post here, BTW)
> 
>                     Name: ml_cloth1.m1v
>    ml_cloth1.m1v    Type: MPEG Video (video/mpeg)
>                 Encoding: x-uuencode

Post a reply to this message

This is fantastic. I've been kicking around the idea of doing a certain
scene with draped cloth for quite some time, but never gave it much thought
because I knew the cloth would be pretty challenging, probably TOO
challenging for me. Therefore I want to encourage you - you're doing great!

It looks like for the first 1/3 of the animation the cloth is actually
stretching to cover the sphere. Or maybe that's what you mean by the vacuum
effect. At any rate, the final appearance is very convincing.

Regards,
Dave

"Michael Langton" <mla### [at] hotmailcom> wrote in message
news:l5knissqgr484uronlseb2i00e18j25i8e@4ax.com...
> Hi.
> Over the last few weeks I've been working on some cloth simulation
> using Hugo Elias' method:
> http://freespace.virgin.net/hugo.elias/
> After seeing the cloth related discussion in pbi, I thought I'd post a
> sample of where I've got up to.
> The actual simulation was done in a C++ program (I did try a
> pov-script one but have now left that for the 5000%+ speedup that
> compiled code gave me...).  The cloth here is a 50x50 grid, with each
> square split into two smooth_triangles based on maximum surface area.
> On the left near the end of the animation you can see my current
> problem - the jagged edges to the shadows caused by it still being
> made of triangles under those nice curves.  I've been wondering
> whether to do it with bicubic_patches instead, but that'll kill my
> nice fast render times (7-8s per 400x300 antialiased frame on a PIII
> 600).
> There is no randomness in this at all currently, hence the
> super-symmetricality of it (it's not quite, actually, because it's not
> centred exactly on the sphere), and the fact that it looks like it's
> falling through a vacuum.
> Well, that's probably enough rambling...comments welcome.
>
> -- Michael Langton (first post here, BTW)

Post a reply to this message

Dave Blandston wrote:

> It looks like for the first 1/3 of the animation the cloth is actually
> stretching to cover the sphere. Or maybe that's what you mean by the vacuum
> effect. At any rate, the final appearance is very convincing.

Maybe the elasticity of the cloth has been set too high

Pabs

Post a reply to this message

So very fantastic.  This kind of thing can't lose in this target
audience.
I noticed a light casts a shadow across the cloth as it first begins to
settle, I almost thought that was some sort of defect.
The wavering interactions at the smaller tighter folds is exhilarating
to watch as well.

Bob

Post a reply to this message

"Dave Blandston" <gra### [at] earthlinknet> wrote in message
news:392ca260@news.povray.org...
|
| It looks like for the first 1/3 of the animation the cloth is actually
| stretching to cover the sphere. Or maybe that's what you mean by the
vacuum
| effect. At any rate, the final appearance is very convincing.

Maybe you were only seeing the light/shadow fall across the top like I
was seeing it?  I had thought it was sucking down onto the sphere
because of the change in brightness there anyway.

Bob

Post a reply to this message

>WONDERFUL, I am quite impress.
>But I don't understand something: You modify the shape of
>your mesh inside povray OR with another soft and you use 
>povray to render this serie of mesh?
The mesh (for every frame) is generated entirely in the external
program.  All my pov script does is #include a file based on the
current clock value - there is a separate inc file for each frame, and
yes, they take up a lot of space, ~1mb each(!).

-- Michael Langton

Post a reply to this message

oh my!

Rick

Post a reply to this message

Very nice! I hope you plan on implementing this using the work by the french
guys I recommended. Their formulas will make it even faster and realistic.
Their documents show the defects of the other method. Keep it up! :) Oh, and
I suggest you give the user the choice of using triangles or bicubics.

Post a reply to this message