animation layers and custom attributes

[ynedelin]

Animation layers and custom attributes?

Has anyone used animation layer in production?

We are on Maya 2010 and running into bugs.

For example animation layers do not seem to like custom attributes.

Also animation does not get refreshed if there is not actual animation
curves at the given frame.

Yury

[Mark Jackson]

Yes, were running layers in production, whilst the animators like it, we have had some issues. If you're using 2010 and characterSets there was an issue that Maya would disconnect the characterSet from it's members, breaking any referencing.... thankfully this is fixed in 2011.

Also it's painfully difficult and slow to get rig of layers once you've used them. For a large scale cutscene, say 4000frms of baked moCap, it can take 20minutes to merge layers down, if it even manages it before crashing. I don't understand how they could have got this process SOOOOOOO slow. If you do the same kind of dense data operation on multiple Trax it takes seconds.

--
http://groups.google.com/group/python_inside_maya

[Paul Molodowitch]

Yeah, there's some serious issues with merging animation layers. We
experienced very frequent crashes and extreme slowness, just as you
did. We were able to at least prevent the crashing by baking the
animation layers using our own command - and the key there is to use
the "simulation" flag when calling bakeResults. My guess is that this
makes a difference because if you don't, maya tries to merge all the
frames "at once", and runs out of memory, while simulation mode does
it frame by frame, freeing up memory as it goes... Just a hunch,
though. Regardless, the underlying call made when you do a merge
layers through the GUI doesn't use "simulation", so I recommend
avoiding that when baking.

- Paul

> --
> http://groups.google.com/group/python_inside_maya

[Mark Jackson]

Thanks I'll give that a go. I was actually thinking of a slightly more radical approach which, in testing avoids all the slowness completely. Basically casting each layers animdata into the Trax as a clip, then simply merging the clips and activating the results. The Trax merge is pretty much instant, no baking involved at all.

--

http://groups.google.com/group/python_inside_maya

[Paul Molodowitch]

Interesting... that may be the way to go, given that bakeResults, even
with the simulation flag, still takes quite a while..

- Paul

> --
> http://groups.google.com/group/python_inside_maya

[Judah Baron]

So you guys are using character sets? We looked at using them a while back and ran into a lot of issues: our reference structure and rig changes make debugging broken connections almost impossible, coordinate system problems transforming animations with "worldspace" affectors, etc, and some similar issues with baking that you all have described here. It looked like the character sets would be safe with something simple, like a skeleton, but we had trouble using them with our rigs. Because of this we stopped. I was disappointed because I think the basic idea is terrific.

--

http://groups.google.com/group/python_inside_maya

[Mark Jackson]

We run fully a referenced animation pipeline and all connections / animations go via the characterSet.... it does need a little babysitting though.

It also depends what version of Maya you tried with chSets. With each release more of the killer bugs get squashed, 2011 has a few specifics in the way the sets are handled and reconnected with referencing. In theory the sets actually make it easier and more flexible for rigging as nothing is name based, unlike normal referencing. All the connections and refEdits go via a placeholder list which is generated from the initial order of the characterSet, broken down into linearValues, angularValuese and unitlessValue lists. The key to it all is maintaining these list orders, referencing just connects animcurveTL to linearValues[122] etc.

Right at the start of our pipeline we wrote a characterSet saver, which enables the riggers to always force the orders of the set to be consistent, after that everything is relatively safe. We get a few failed files from time to time, but most of them we can recover, and it's normally as a result of an updated source rig with a duff chSet order in the first place. Some of the GoldenEye cutscenes are huge, 5000frms, all dense MoCap data, 10-15 full rigs + props and all referenced. If you can tame it, it's a great production time saver.

I do have to say that the initial love that the guys had for AnimLayers has disappeared, and all of a sudden they're running back to Trax due to bugs in the way the layers are handled.... and the lack of decent methods to remove the damn things once made!

Mark
 

--

http://groups.google.com/group/python_inside_maya