transferring low res syflex to high res geo

[Kris Rivel]

I'm curious how others handle layered or very high res geo clothing?  I have a pretty high res geo of layered clothing on a character.  I'm getting some decent results with just applying syflex to the geo, enabling self collision, boosting the subframe steps, etc. and letting 'er rip but its pretty slow.  I get fantastic results using one, lower res proxy mesh.  Whats the best way to transfer or copy a low res sim to a high res mesh?  I was creating nulls on each point on the lower res geo and enveloping the high res too it but now I'm dealing with managing almost 2k worth of nulls and any envelope corrections are impossible.  Any other suggestions other than brute force?

Kris

[Eric Thivierge]

Use ICE to get the closest location on your low res (static, undeformed copy) from the high res mesh and store the offset transform from that location. Then use the reinterpret location onto the deformed low res geo.

--------------------------------------------
Eric Thivierge
http://www.ethivierge.com

[Orlando Esponda]

Have you tried Cage Deformer?  Also, storing a location with ice, and then reading the location on each frame.

Just a couple of ideas...

On Thu, Dec 6, 2012 at 8:58 PM, Kris Rivel <kris...@gmail.com> wrote:

I'm curious how others handle layered or very high res geo clothing?  I have a pretty high res geo of layered clothing on a character.  I'm getting some decent results with just applying syflex to the geo, enabling self collision, boosting the subframe steps, etc. and letting 'er rip but its pretty slow.  I get fantastic results using one, lower res proxy mesh.  Whats the best way to transfer or copy a low res sim to a high res mesh?  I was creating nulls on each point on the lower res geo and enveloping the high res too it but now I'm dealing with managing almost 2k worth of nulls and any envelope corrections are impossible.  Any other suggestions other than brute force?

Kris

--
IMPRESSUM:
PiXABLE STUDIOS GmbH & Co.KG, Sitz: Dresden, Amtsgericht: Dresden, HRA 6857,
Komplementärin: Lenhard & Barth Verwaltungsgesellschaft mbH, Sitz: Dresden,
Amtsgericht: Dresden, HRB 26501, Geschäftsführer: Frank Lenhard, Tino Barth

IMPRINT:
PiXABLE STUDIOS GmbH & Co.KG, Domicile: Dresden, Court of Registery: Dresden,
Company Registration Number: HRA 6857, General Partner: Lenhard & Barth
Verwaltungsgesellschaft mbH, Domicile: Dresden, Court of Registery: Dresden, Company
Registration Number: HRB 26501, Chief Executive Officers: Frank Lenhard, Tino Barth


--
Diese E-Mail enthält vertrauliche und/oder rechtlich geschützte Informationen. Wenn Sie nicht
der richtige Adressat sind oder diese E-Mail irrtümlich erhalten haben, informieren Sie bitte
sofort den Absender und vernichten Sie diese Mail. Das unerlaubte Kopieren sowie die
unbefugte Weitergabe dieser Mail ist nicht gestattet.

This e-mail may contain confidential and/or privileged information. If you are not the intended
recipient (or have received this e-mail in error) please notify the sender immediately and destroy
this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is
strictly forbidden.

[Ed Manning]

ICE?

I'm pretty sure there's either a sample scene or a tutorial floating around online on how to deform one mesh onto another. 

It's a little late in the day for me to actually work this out, but basically, you'd have a highres mesh with an ICE tree getting position deltas (change from last frame to current frame), interpolated, from the corresponding locations on the low-res mesh.  Might want to have matching UV's, but that might not be necessary. 

I guess really it's a cage deformer running in ICE, but without any kind of static weight binding from one mesh to the other.

Or I suppose you could try the old cage deformer, but that's horrible -- slow and inflexible.

Or extract a lot of curves from the low-res mesh, plot their shapes, and envelope the high-res to them, if you need something like your current workflow?

Or maybe I'm not understanding how complex the geo is, but maybe you could cache the low-res simmed mesh, then extract the poly clusters that correspond to the different parts of the high-res clothing.  Then either leave the extractions live or if it's too slow, cache them. Then apply however much subdivision and editing of the geo you need to the cached extracted meshes.  Essentially leaving the modeling stack live on top of the low res sim.

But that's getting kind of silly, I think.  ;-)


Ed

[Ed Manning]

wouldn't the stored offsets be in world space?  Wouldn't you want them in tangent space to not crumple weirdly?

[Kris Rivel]

Thanks for the suggestions.  Eric...your idea is what I had in mind but having a hard time figuring out the right node usage once I extract closest point.  How do I store an offset?  Cage is ridiculously slow so thats out.  

[Eric Thivierge]

Well yes you're going to have to figure out the transform space when storing in the point's local space. Though I think you could hack your way through it by ripping apart the stick to location node.

--------------------------------------------
Eric Thivierge
http://www.ethivierge.com

[Matt Lind]

I wrote an article on this technique many years ago for HDRI 3D magazine:

 

Assuming Syflex works on NURBS surfaces and your cloth is a planar polygon mesh deformed with wrinkles and whatever.  I would surface constrain the mesh to a NURBS surface of similar size/shape, and run Syflex on the NURBS surface.  When you’re testing the simulation, you can hide the mesh to get the performance.  When the simulation is to your liking, unhide the mesh to check the results.  When you have something you really like, you can optionally plot the results as shape animation onto the polygon mesh.

 

If you prefer, you can also plot out as envelope animation by making a bunch of nulls to act as deformers of your polygon mesh, and then constrain them to the control points of the NURBS surface.  Plot transforms for the nulls whenever you feel the results are to your liking.  If you want multiple ‘takes’ to choose from, you can plot to a mixer source instead of directly to the Transform parameters as FCurves.  You can then play with the mixer sources to further massage into a result you like.

 

This would give you a lot of freedom and quick iteration.

 

Matt

 

PS – if Syflex doesn’t work for NURBS, you can use a low res mesh as the working copy.

[Eric Thivierge]

Kris are you using the standard Cage deformer or the ICE based "Hull Deformer" setup?

--------------------------------------------
Eric Thivierge
http://www.ethivierge.com

[Kris Rivel]

Thanks Matt...could give that a go too...although I think ICE syflex only works on poly?  Eric...good suggestion for a starting point..will try that.  I was referring to the default cage deformer...didn't think about hull deformer.  I was looking at a few images and demos on syflex's website for the Assassin's Creed trailer.  Pretty awesome setup they had with multiple layers of clothing.  The ICE Syflex seems much more powerful at first and is definitely faster I think but lacks some of the tools the addon version has like weightmaps.  There's another one I'm looking for an answer too.  Although I figured if I was able to blend a syflex mesh with a weighted mesh using a weight map in ICE, I'd be set!

[Simon Anderson]

We would take the high res geo, reduce it, then run a custom script that generated nulls that were cluster constrained to the points on the geometry and simulate it, then envelope those points onto out the high mesh. This was one of our work arounds, depending on the shot and what was needed, but this solution seemed to work pretty well, as long as your create the geometry from a default flat position and then simulate into position.

we tried cage deformers on some things, but found the size of the cage deformers to be huge compared to the enveloping.

I think each situation calls for different approach. Found that some times you would get great results with one method and then the same method would create horrid results.

--
-------------------
Simon Ben Anderson
blog: http://vinyldevelopment.wordpress.com/

[Ed Manning]

that's why I was thinking UVs might help -- use them to interpolate the spatial offsets?

[Kris Rivel]

Thanks Simon...I was using that method as well.  It works fine on some other elements but for this one particular case, it was difficult due to the layered fabric.  Will give that another go as well.  Didn't think about doing a poly reduction, keeping it live and using that as my null deformer group...might be more manageable.

[Ed Manning]

Anyway, past my bedtime.  Kris, you're in much better hands than mine if you listen to Eric, Matt et al.  Best of luck with it.

[Matt Lind]

You’d have context switching issues doing that as each vertex has multiple samples (UVs).

 

Matt

[Matt Lind]

EDIT:

 

Surface constrain the nulls to the NURBS surface, not to the control points.  This would allow you to drive the orientation of the nulls as well as translation.

 

Should be able to do the same trick with a low resolution polygon mesh as the simulated surface if NURBS doesn’t work.

 

Matt

[Eric Thivierge]

Locations have the PolyReferenceFrame or PointReferenceFrame that you could use which I think the Stick to Location method / Hull Deformer uses...  

--------------------------------------------
Eric Thivierge
http://www.ethivierge.com

[Matt Lind]

They’ll stick, but they won’t necessarily have a desirable or easily predictable orientation as the results will largely be dependent on the structure of the mesh topology.

 

Matt

 

 

 

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Eric Thivierge

Sent: Thursday, December 06, 2012 8:06 PM
To: soft...@listproc.autodesk.com

[Raffaele Fragapane]

With the right topology and density, they will.
We bound millions of points worth of creatures between dragons and dinos that way across multiple resolutions ;)

[Eric Thivierge]

I never said it was going to be perfect or easy. :) Maybe averaging the neighboring RefFrames would give a better result?

--------------------------------------------
Eric Thivierge
http://www.ethivierge.com

On Fri, Dec 7, 2012 at 3:16 PM, Matt Lind <ml...@carbinestudios.com> wrote:

[Matt Lind]

As a user my main worry would be editing one of the mesh topologies would unexpectedly cause some nulls to change orientation because the reference frame changes.

 

A safer approach would be to create a smooth and continuous coordinate space and align the nulls using that data. If the mesh topology is edited, the nulls won’t have the rug unexpectedly pulled out from underneath.  Nvidia has sample code to compute such a space for normal maps (NvMeshMender?).  The code could be modified to output a coordinate space as a weightmap, vertex color, UVspace, userdatablob, or whatever suits the purpose.

 

Matt

 

 

 

 

 

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Eric Thivierge
Sent: Thursday, December 06, 2012 8:20 PM
To: soft...@listproc.autodesk.com
Subject: Re: transferring low res syflex to high res geo

 

I never said it was going to be perfect or easy. :) Maybe averaging the neighboring RefFrames would give a better result?

[Orlando Esponda]

Maybe I'm missing something here, because the solutions are rather complex  hehe  so please bear with me if I'm totally out of context.   Here's a little video showing how my first suggestion works. The smoothing part, well, that's up to you, I was just playing around, but who knows, maybe it works for you...

http://www.screencast.com/t/jZIJ3L3lEDPc

Orlando.

[Leonard Koch]

This compound creates pretty good results and is fairly versatile. 

I haven't run into a cloth skinning scenario where this approach wasn't sufficient, so I hope this will help you out.

I have included a sample scene.

Enjoy

[Eric Thivierge]

Why are we even talking nulls when Kris already said it was getting too cumbersome to use nulls? 

--------------------------------------------
Eric Thivierge
http://www.ethivierge.com

On Fri, Dec 7, 2012 at 3:33 PM, Matt Lind <ml...@carbinestudios.com> wrote:

As a user my main worry would be editing one of the mesh topologies would unexpectedly cause some nulls to change orientation...

[Raffaele Fragapane]

Why are we even talking at all when Leonard pointed out there's a factory node doing exactly what is being asked for? :p

[Leonard Koch]

That is not a factory node.

I have built similar systems in the past.

So it didn't take me too long to make a proper compound for this.

[Matt Lind]

You suggested using a reference frame.  I merely pointed out a potential trouble spot.

 

 

Matt

 

 

 

 

 

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Eric Thivierge
Sent: Thursday, December 06, 2012 8:58 PM
To: soft...@listproc.autodesk.com
Subject: Re: transferring low res syflex to high res geo

 

Why are we even talking nulls when Kris already said it was getting too cumbersome to use nulls? 

[Raffaele Fragapane]

Oh ok, thought I was just behind on features (haven't used syflex since Sucker Punch).
Good stuff :)

--
Our users will know fear and cower before our software! Ship it! Ship it and let them flee like the dogs they are!

[Ben Davis]

And merry Christmas to you too Leonard, I have a scenario that this compound will find usage in right this morning.
Thanks for sharing!

Ben

--

Benjamin Clifford Davis

www.moondog-animation.com 

office:   +33 9 50 04 76 15
mobile: +33 6 88 48 54 50

6 bis avenue des Iles
74000 Annecy
FRANCE

[olivier jeannel]

Loved that one ! Would be worth uploading this to vimeo/Ice group. Excellent reference imho :)

Le 07/12/2012 05:36, Orlando Esponda a �crit�:

Maybe I'm missing something here, because the solutions are rather complex� hehe� so please bear with me if I'm totally out of context. � Here's a little video showing how my first suggestion works. The smoothing part, well, that's up to you, I was just playing around, but who knows, maybe it works for you...

http://www.screencast.com/t/jZIJ3L3lEDPc

Orlando.

On Thu, Dec 6, 2012 at 10:19 PM, Eric Thivierge <ethiv...@gmail.com> wrote:

I never said it was going to be perfect or easy. :) Maybe averaging the neighboring RefFrames would give a better result?

--------------------------------------------
Eric Thivierge
http://www.ethivierge.com

On Fri, Dec 7, 2012 at 3:16 PM, Matt Lind <ml...@carbinestudios.com> wrote:

They�ll stick, but they won�t necessarily have a desirable or easily predictable orientation as the results will largely be dependent on the structure of the mesh topology.

--
IMPRESSUM:
PiXABLE STUDIOS GmbH & Co.KG, Sitz: Dresden, Amtsgericht: Dresden, HRA 6857,

Komplement�rin: Lenhard & Barth Verwaltungsgesellschaft mbH, Sitz: Dresden,
Amtsgericht: Dresden, HRB 26501, Gesch�ftsf�hrer: Frank Lenhard, Tino Barth

IMPRINT:
PiXABLE STUDIOS GmbH & Co.KG, Domicile: Dresden, Court of Registery: Dresden,
Company Registration Number: HRA 6857, General Partner: Lenhard & Barth
Verwaltungsgesellschaft mbH, Domicile: Dresden, Court of Registery: Dresden, Company
Registration Number: HRB 26501, Chief Executive Officers: Frank Lenhard, Tino Barth


--

Diese E-Mail enth�lt vertrauliche und/oder rechtlich gesch�tzte Informationen. Wenn Sie nicht
der richtige Adressat sind oder diese E-Mail irrt�mlich erhalten haben, informieren Sie bitte

[Kris Rivel]

Leonard...you are AMAZING!!!!  I owe you a few round of beers...this is great...fast and super easy.  Much thanks!!

[Kris Rivel]

Thanks for the vid Orlando..had no idea it was that simple...really great.

[Leonard Koch]

I'm always hire-able to quickly solve problems like this.

When you need a small production bump like this out of the way, just shoot me an email and then we can see what needs to be solved, so that you don't waste too much production-time and thereby money.

Apologies if this sounds overly commercial guys. It's just always nice to help people and have some small side jobs.

[Kris Rivel]

Hey thanks!  I'm definitely interested in that.  Let me sit on it a bit and see how this project goes.  I may take you up on that though.  I honestly wonder if there might be an opportunity to make a little money with something like this.  I'm not much of an ICE guru yet but turning into a pretty good business man :-)  Been a soft generalist for 12 years but turning more to a business person and less of a day to day artist.  Just thinking out loud.  Not sure if that would interest you or not.  Where are you located?  I would definitely like to keep in touch though and will most definitely use you on ICE related stuff..you've proven you know more about it than I do.

[Leonard Koch]

Good to hear.
I'm situated in Hamburg, Germany.

[Kris Rivel]

Hi again...any reason why it sometimes will seem to fail a frame or two in and just stop following the referenced mesh?  I have two meshes...one is following fine, but the other fails quickly.  You see it move a bit, then stops....its a very slow, gradual movement so its not a sudden change or shift.  I tried turning up the sim steps but doesn't effect it.  Just curious if you know what might cause that.  The two surfaces are very close and almost match...just one has slightly lower topology resolution.  It seems to work fine in the original scene its applied to...but if you import it as a model, it breaks.  Sometimes...if I just select the syflex mesh and play the scene...it works!  But if I have nothing selected or if I select the mesh with your compound...it breaks :-/

Kris

[Kris Rivel]

Sorry...that was meant for Leonard :-/

[Leonard Koch]

Sounds like a more complex ice problem that might have nothing to do with the compound itself.

I'd have to take a look at the scene. Maybe you could strip it down to it's bare essentials so that the list can take a look?

Unless it falls belows XSI's threshold of precision, - which I'm sure it is not - the speed of the movement shouldn't have any impact on the functionality of the compound. It is just smaller vector differences being computed. It shouldn't make any difference.