PDA

View Full Version : Metalink Morph Symmetry Issue



Chris Jones
10-23-2012, 09:07 PM
Here we have some symmetrical fibers parented to a symmetrical morphing object. I applied ClothFX to the morphing object and scanned the motion to create an mdd, then applied FX_MetaLink to the fibers. As you can see, the fiber symmetry is not retained. FX_Hardlink causes the problem to flip to the other side.

http://forums.newtek.com/attachment.php?attachmentid=108736&d=1351049744

108729

I'm not experienced with ClothFX so it's possible I'm misusing it in this instance. Any thoughts?

Thanks!

(c:

nickdigital
10-24-2012, 01:54 AM
I say fog it and send the content to NewTek. I though it might be a point order issue but the fibers don't look to be deforming all wacky.

RebelHill
10-24-2012, 02:15 AM
Its just meta-link... not very good.

Chris Jones
10-24-2012, 07:26 AM
Fogged, thanks.

I'm starting to think it might be something to do with the morph object, as I can't seem to reproduce this with a fresh new mesh. In the meantime I guess I'll be needing another way to lock the fibers to the skin without having to morph the fibers themselves (ie match them individually to the morphs in the other object (ie painstaking)).

(c:

RebelHill
10-24-2012, 09:16 AM
Oh, chris.... its u. I didnt even register the username before.

Well... here's what I do.

I tend to just ffx directly on mesh in layout, no guide geo, as having it match with morfs etc is a huge pain. Using the layout ffx tool/edit guides etc is easier and sticks properly to deforming geo. The workflow issue here is working with subD stuff... as you cant go switching subD levels after styling. To get round this, I either have a master copy character/rig, with the subdiv/hair done, and merge motions back onto it fro a bald copy... or... I create a second geo of just the cut out hair bits, which I scale down along normal the tinyest amount. This geo ofc being copied from the original head geo. This can then be styled, locked, made fully transparent, and its subD left alone, and it'll still follow with the main mesh morfs, bones, etc.

The other alternative if this isnt pressing right now... is jsut to wait for 11.5. That's gonna have layout guide styles stored in vmaps in the objects themselves, so the whole problem described above will go away.

Btw... that eyelid deformation u were having issues with a while back... I ofc recalled the answer recently when I hit on something similar... there was nothing wrong with ur nodes or LW, it is ofc displacement order. Had u set it to before world, itd work fine.

Chris Jones
10-24-2012, 06:14 PM
Thanks Rebel,

I wasn't able to get enough accuracy out of the Layout FFX for things like eyebrows, and yeah the inability to change subD levels is a big deterrent. Also having the fiber strands as a separate object seems like a tidier and more future-proof approach, if only they could be properly glued back on... Maybe having the styles stored with the object will be the go (especially if there's a way to edit the vmaps).

As for the eyelid deformation, I just checked the problem scene and node displacement is already set to before world - the displacement strays no matter what the setting. I ended up going with effectors anyway, which holds up provided I don't go super closeup on the eye.

(c:

Chris Jones
10-24-2012, 07:06 PM
I can now confirm that the issue isn't isolated to the objects in my example, as I just recreated it with a new scene and new objects.

RebelHill
10-25-2012, 01:10 AM
I can now confirm that the issue isn't isolated to the objects in my example, as I just recreated it with a new scene and new objects.

Yeah... metalink it is. Its always been reasonably poor if the target mesh isnt already a VERY close match to the source. Ive tried metalinking dynamic wigs before to things brought through from SI, and strands wil "ignore" a hair that its 1mm from, and stick to one 5mm away... then jump back again, etc. Used on larger shapes, with looser geo... its failings often arent noticable, but try to get precision out of it and it falls flat. There was briefly a metalink node version... but that sucked real donkeys. Word on the street is its making a comeback soon.. So I guess we'll have to wait and see how it works out.

Just for the record though... I can build a "metalink" in tool in ICE in less than a minute, and its PERFECT... so it cant be that bloody hard.

Chris Jones
10-25-2012, 05:38 PM
Here's hoping for that metalink node then. Or ICE for LW... :cool:

erikals
12-18-2012, 10:16 PM
hi, were you able to improve it?
i found that softFX and metalink with "Smoothing" on worked ok, not super, but ok, sorta...

erikals
12-19-2012, 11:29 AM
also see > http://forums.newtek.com/showthread.php?132336-Eyes-and-Wrinkles&p=1288237&viewfull=1#post1288237

erikals
12-21-2012, 02:46 PM
anchor might work, though tedious...

(just tested it on 2-3 guides)

erikals
12-21-2012, 03:41 PM
Metalink Guides, testing some stuff...
http://youtu.be/-aoBTCAR0Cg

edit:
nah, looks like there is a problem getting a metalink to drive a metalink... [ugh!]

erikals
12-21-2012, 04:24 PM
The other alternative if this isn't pressing right now... is just to wait for 11.5. That's gonna have layout guide styles stored in vmaps in the objects themselves, so the whole problem described above will go away.

xrossing fingers for that feature... :goodluck:

erikals
12-21-2012, 06:00 PM
Metalink Guides, video 1...
http://youtu.be/-aoBTCAR0Cg

Metalink Guides, video 2... interesting result :]
http://youtu.be/qaYh_FDMWvs

Chris Jones
01-09-2013, 07:30 AM
Thanks for your experiments with this erikals, looks like a possible workaround in the event that there are no forthcoming metalink improvements. Not sure if it will be enough for eyelashes though, metalink does not like them at all.

erikals
01-24-2013, 09:54 AM
hi, forgot to answer this :]
it's cool to see that more people would like to see these challenges solved.

this said though, bad news, i just tested something i thought could work, but nope :/
didn't work...

got closer, but it's time consuming and might have a bug... (see attachment)
http://youtu.be/x02J4tNo_4g

it does seem to work excellent in the example you posted though...
http://www.youtube.com/watch?v=PNiuRnisK98

erikals
01-24-2013, 11:00 AM
btw, have you tried hardlink instead of metalink? might work...

Chris Jones
01-26-2013, 03:41 AM
Hardlink has more trouble handling these deforming surfaces than metalink, unsurprisingly. My example works (from a distance at least), because I had to resort to weighting and morphing each eyelash to match the eyelids, which of course is not very practical (nor accurate).

VermilionCat
01-26-2013, 04:33 AM
Dpont's Meta Pointer seems to work. You have to bake the cage objects though...

erikals
01-27-2013, 10:03 AM
what problems did hardlink give though, same thing?... inaccuracy?...

hardlink example >
http://forums.newtek.com/showthread.php?133007-How-to-make-objects-stick-to-uneven-surface&p=1295071&viewfull=1#post1295071

dpont
01-27-2013, 10:40 AM
...Meta Pointer seems to work. You have to bake the cage objects though...

Don't need to bake with Meta Fit node.

Denis.

erikals
01-27-2013, 10:57 AM
That could be interesting! \:]

hmm, so it could be used as a lattice deformer as well?...

VermilionCat
01-27-2013, 11:45 AM
Don't need to bake with Meta Fit node.

Hi, Denis!
Meta Fit works similar to Metalink in my test. MetaPointer works completely symmetrical.

erikals
01-27-2013, 11:48 AM
might be a misunderstanding, as the word should have been "calculate" and not "bake" ?

VermilionCat
01-27-2013, 11:51 AM
Yeah, sorry. You have to generate .mdd for the cage object.

dpont
01-27-2013, 12:12 PM
Hi, Denis!
Meta Fit works similar to Metalink in my test. MetaPointer works completely symmetrical.

Meta link works symmetrical too, no mdd,

110718

Denis.

VermilionCat
01-27-2013, 12:49 PM
Then am I doing something wrong?? Now they seem all different...

dpont
01-27-2013, 01:02 PM
Then am I doing something wrong?? Now they seem all different...

Don't know,
Meta Pointer and Meta Fit node use the same internal process,
so I can't explain such different result.

Denis.

Chris Jones
01-27-2013, 05:26 PM
what problems did hardlink give though, same thing?... inaccuracy?...A lot more drift, with the hairs moving away from each other in clumps.


Dpont's Meta Pointer seems to work. You have to bake the cage objects though...Can't seem to get it to do anything... how do you use it?

VermilionCat
01-28-2013, 12:14 AM
Chris, Denis is right. MetaFit works symmetrical in your scene without mdd.
Now I don't understand why my little scene doesn't work...

dpont
01-28-2013, 01:34 AM
..I don't understand why my little scene doesn't work...

May be because your mesh is just flat quad,
I can get it working if it is triangulated.

Denis.

VermilionCat
01-28-2013, 02:07 AM
Thanks, Denis! It works that way! You are great man!

lino.grandi
01-29-2013, 07:52 AM
When metalink it's already applied and you make some changes to the objects, my advice would be to take out Metalink and apply it again. In a situation like the one you're showing, it should work perfectly. And yes, using tris is very important.

Chris Jones
01-29-2013, 06:58 PM
But triangulating isn't really a solution when you need SubDs... reapplying metalink doesn't restore the symmetry either.

My findings are that metafit is more accurate than metalink, but it's still not accurate enough for eyelashes (not to mention that it's now crashing my main scene even when trying to delete the node, but that's another story...).