[Sidefx-houdini-list] Holding Bone Weights
ld at dneg.com
Wed Jun 13 07:20:59 EDT 2007
> technically that is what the second input on the editSOP is supposed
> to do. It should be creating a local coordinate system on the deform
> and on the rest and then use a quaternion to move the geometry around.
> Thing is in practice the more the deformation changes the less likely
> the editSOP will be valid. I think of it like a noise field.. the
> noisier the field the less likely you will get a wide range of use
> out of the blendshape. strictly speaking that's what we used for
> character finaling on the wild. the problem is no matter what they do
> with the coordinate system it still gets a bit wonky in certain
> deformations over time. I was intrigued with the radial basis
> functionSOP Simon put out, but it can take a while before you get a
> strong relationship between the mapping points and the changes you want.
> for rigging on the other hand this system should be able to give you a
> higher percentage of success when you use it for bone angles and the
> like. the reality is however, that any deformer after a deformSOP is
> slow, so you avoid it. Which means you can use the createeditsop
> hscript command to generate a preDeformSOP editSOP result so you can
> do use it in a blendSOP above the deform. Again, you can only get so
> many in before you start paying the price.. in memory and performance.
> you idea would be helpful if it was done in the deformSOP. Anything
> else isn't as fast or nice once you start getting complicated.
Yeah, that's exactly what I'm talking about. Layering SOP deformations
is cool but slow.
I've been asking the guys at SESI to consider adding this option to the
deform SOP so that it could work in the same way as say the Lattice SOP
on a per-bone basis so that only bones that have been explicity moved
have an effect. A bone that has moved simply because it's in a hierarchy
would not have any effect.
I love the oriented editing available through the edit SOP via the Morph
tool etc. It's a life saver. It is a blendshapes approach however so
there are some restrictions there. The other methods you talk about are
all valid and very cool and I've used/looked at them all (thanks again
Simon for the gooviness). We've put together various other flavours of
this on different occasions. It's good to see SESI put in the Attribute
Orient SOP too.
You can access metaball transforms in VEX which is cool.. :-)
Go on SESI, you know you want to! 8-)
> we have implemented a nice rig system now with preBlendshapes,
> wiredeformer, then a deformSOP. when used correctly it is very nice to
> control the deformation and still give the animators the freedom.
> Thing is it can be slow. I know about the assume coordinate changes
> toggles (blend and deform), the morph mode in the blendSOP etc. The
> blendSOP will cook only those inputs with values above 0, but you
> still pay the price for numbers here and you are bound to get multiple
> floats in there somewhere. the morph mode helps since only the deltas
> are stored in the editSOPs once the file has been saved after you
> toggle it on and reload the file, so the memory hit is less. The wire
> deformer isn't optimized with the assume coordinate changes so that
> isn't much help either. the difference in attributes also can be a
> pain as well, but you deal with it.
> I haven't played enough with the metaball system yet to know if it is
> potentially easier to get more out of the rig. I saw the videos Calin
> put out. It looked great, but I was worried about the workflow quirks
> you had to do to set it up.
> Louis Dunlevy wrote:
>> I'd be curious to see what people think of this idea of a "relative"
>> deform for Houdini bones/metaballs. Any takers?
>> Louis Dunlevy wrote:
>>> I was going to suggest the spreadsheet but you're there already I
>>> see. A bit messy though. I quite like the sliders on the bones too.
>>> For facial rigs have you had any joy with the Morph tool giving it a
>>> rest position? When you hit the "Save Deltas" button it will do an
>>> Attribute Orient style deform based on what you've done in your edit
>>> SOP. I'm sure you're already aware but just in case..
>>> I really would like to see a new type of bone behaviour like the
>>> "relative" mode in Maya's cluster deformer. That lets you throw in
>>> tons of extra deformers that have no affect whatsoever until their
>>> transform channels have changed. I find the problem with Houdini's
>>> bones is that they *always* have an effect on the geometry whether
>>> you like it or not. I think it's a big issue for facial rigs and
>>> secondary deformers.
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
More information about the Sidefx-houdini-list