Jump to content
Musetatron

AAA Production quality character animation in houdini?

Recommended Posts

i still think that 3 main branch of people have to live together in CG

- the artist , who hate tons of buttons everywhere , and who essentially want to focus on pretty pictures and create in the most straight forward way

- the tds, who love technics and art and who really love to go down to the metal of the machine to build impressive assets and pictures

- the R&D guys who just focus on the API and the workflow and want to bring the best tool ever.

In CG talk i've said previously that imo the best way to solve this for sessi would be to split houdini in 2 package with the same core but with a very different UI.

Houdini Artist => would be a tool very simple and very user friendly which would only be a place to interact with OTL

Houdini Core => would be the toolbox where TD and R&D guys assemble tools for artist

I agree with your categories of different disciplines but it is somewhat constricting to box a person into one role. People work differently, we all have different strengths and weaknesses, likes and dislikes. Some, given the freedom to do what they want can blur the line between technical and artistic. The task of making Houdini friendly to artists yet retaining it's technically open foundation is not a futile effort in my eyes.

One of the core reasons why Maya was so successful was it could marry the two into one package. There was artistic freedom with a large degree of technical flexibility in the form of mel and the C++ api.

To split the package would have a disastrous effect i feel not just for the userbase, but also the time/money spent doing so for SESI.

Again in my opinion.

Edited by abvfx

Share this post


Link to post
Share on other sites

Well maybe you're right Andrew. But do you think that SESI could accomplish the miracle of

- keeping houdini strength and fantastic sentiment of freedom

- pack this in a very easy intuitive tool that will have as much UI sugar candy as modo for ex ?

 

This is also a huge task ! But i would not consider

- Houdini Artist

- Houdini Core

2 separate app.

 

it would only be one app but with 2 very radical presentation.

- imagine more a librarie of OTL that represent the tools and that are easily expose

- the OTL complexity would be hiden in H Artist while H core would contain all the network.

 

I agree that the goal is not to develop 2 apps at the same time. But to develop an app than can be layout in 2 different radical way.

- one with Shelf / Shortcuts / intuitive interaction

- one with all the brute force power we love in H

 

But i absolutely agree that developing 2 apps is a waste of time.

Edited by sebkaine

Share this post


Link to post
Share on other sites

This sounds a little clearer that probably what i had in my head.

I still think it is better kept as one unified pieced but what you are suggesting is something we all know SESI is working on. The new hair tools for example. They aren't ready for prime time imo but they are on the right track.

SESI have more autonomy than most companies out there, while they still market Houdini is primarily a FX package because of years of invested development and interest in that area, they are branching out more and more with each release, capturing audiences of other packages as they go. Yet they always find a way of keeping it "Houdini". I think we are going to see more and more "UX" features in the future.

So I think time will sort this out.

Share this post


Link to post
Share on other sites

I agree that thing are progessing in the right direction. But i think the gap beetween the actual houdini and an houdini app that can be grab as a main tool in place of Max / Maya / XSI

is still monumental. Each time i try to push houdini outside the FX area in the place i work, each time people smile and and tell me that houdini is too complex and only for nerds.

 

I don't need so much update from an FX / Lighting / Proceduralism POV , i love houdini the way it is.

I could only complain about Viewport and Real Time feature that are not as good as i want them to be.

 

But to put Houdini user experience on a higher level than AD tools , will not ask only small new UX additional tool.

The modification must be seriously radical en embrace a new paradigm imo.

 

i think you are right in keeping things unify but i do think that some radical modification are needeed in

- viewport

- connecting texture

- grab channel / set keys

- uv layout

- modeling / sculpt area

- animation workflow

and those radical modification must not destroy what we all love in H , "the freedom to do everything we want"

 

Thus having this kind of UI pattern in Artist Mode / TD Mode is a good idea imo if Houdini want to compet on other area than FX.

But maybe SESI just want to be the best FX tool provider and don't spread in other area, which is not bad after all.

 

But today i simply can't work only in H, Maya still has some very good point for him especially ... the Viewcube (this one is a joke in case you were ready for stoning ...)  :)

Edited by sebkaine

Share this post


Link to post
Share on other sites
Guest mantragora

... the Viewcube ...

1. Learn C++.

2. Go to HDK Viewport section about Hooks.

3. Write custom SceneHook.

4. Profit!

  • Like 1

Share this post


Link to post
Share on other sites

AD Viewcube in houdini ... the dream is coming  true ! :)

Edited by sebkaine

Share this post


Link to post
Share on other sites
Guest mantragora

I was seriously considering making one for 1 April :D

Share this post


Link to post
Share on other sites
But to put Houdini user experience on a higher level than AD tools , will not ask only small new UX additional tool.

The modification must be seriously radical en embrace a new paradigm imo.

 

i think you are right in keeping things unify but i do think that some radical modification are needeed in

- viewport

- connecting texture

- grab channel / set keys

- uv layout

- modeling / sculpt area

- animation workflow

 

basically all things that are non precedural.

I think even if it's a procedural software, it would hugely benefit having a good interaction even for a TD person.

If i need to draw a couple of curves and tweak the vertices and shape like i would do in any other software, doing this in houdini is just painful and slow, so much that the choice is often to do that small task somewhere else, and import back the geo in houdini.

Edited by MENOZ

Share this post


Link to post
Share on other sites

Well we had a few animators working in houdini over the last 6 months. None of them touched it before and they were mix of maya and max users previously. To my big surprise, the interface was the least of the worries. Actually not a single person complained about it because it's so customisable, it just takes a few minutes of setup to make the UI simple enough for animators. The real issues we are having all come from 3 areas. I think this is a very productive discussion now so I'll list them here. I'm not having a rant at all, just stating our findings so far.

 

1. Viewport speed and stability. I had to radically change the way I like to make rigs for animator for them to be usable in terms of speed. Good example is transparent control objects. Animators love them. Being able to grab forearm geometry instead of bone or null feels very organic and it's easy to setup. However it kills the rig speed instantly in H14 (or the transparency on the bones in general). The same goes for skinning and most of the deformers. I wouldn't have animators work with all the heavy deformation rig in maya either, but for preview it's great to see what mesh is doing once in a while. Is it slow in maya if you turn on heavier meshes, of course, but houdini comes to standstill. Textures, no chance for good speed. We are getting decent speed right now, when we use the simplest proxies and only animate bones and nulls, but put more characters in and you're in for a trouble.

 

2. Animation reuse. Almost unthinkable right now. Copying keys between scenes, crash. Exporting and importing .chan/.chn files, impossible to get the import correctly back unless you are super precise with scoping the channels. Presets work most of the time, but then your preset list is really, really long after 20-30 shots. (And we'll be outputing 200 a month come next january.) The only way around that right now would be to write fully custom anim library kind of thing, but frankly we simply don't have time to deal with that any time soon. 

 

3. Digital asset workflow. As much as I really love it for most of the pipeline steps. It's very cumbersome for standard animation shot work. Yes if you have enough time to be doing really meticulous layout then it's fine, but in really fast turnarounds it tends to be mess and very often you just need to override one or two things in rig, environment, etc. but still need to keep the asset live to get potential updates. This one I don't know how to solve. Mainly because after last few months I have a phobia from updating rigs in the scene to their new version. Always 50:50 whether it crashes or not.

 

 

And outside of these it's simply overal stability. We are having without any exaggeration 10-20 crashes a day when animating. Whether it's moving keys, working in animation editor, updating rigs, scrubing the timeline, you name it. I'm certain many of them come down to wrong drivers, wrong graphic cards (we're all on GTX770 and 760), wrong rigs but it seems like awful lot of them anyway.  Quite frankly I'm getting a bit tired of bug reports, simply because it is starting to be very difficult to narrow the problem down so we can post a clear reproducible bug.

 

Edit: just noticed a few meaning changing typos. Of course I meant wouldn't have animators work with heavy deformation rig in maya. I'm not a monster :)

Edited by Milan
  • Like 3

Share this post


Link to post
Share on other sites

That's a very interesting post Milan. Real Production experience testimony in Rig / Anim on H are extremely rare ...

From houdini architecture itself the canvas to build a killing animation package is there.

But there are imo too much risk / trial / randomness to build a full animation pipeline on H until SESI decide to push hard in that direction.

 

Pyside UI is a BIG first step in that direction.

Share this post


Link to post
Share on other sites

Yeah, thanks Milan! Make sure you submit this to SideFX as an RFE, this info would be greatly appreciated!

 

I've seen similar frequency of crashing while animating, and it's hard to submit bug reports because they are often not reproducible. Hopefully the automatic stats collecting + reporting will help make it easier to find and fix those bugs.

Share this post


Link to post
Share on other sites

Well we had a few animators working in houdini over the last 6 months. None of them touched it before and they were mix of maya and max users previously. To my big surprise, the interface was the least of the worries. Actually not a single person complained about it because it's so customisable, it just takes a few minutes of setup to make the UI simple enough for animators. The real issues we are having all come from 3 areas. I think this is a very productive discussion now so I'll list them here. I'm not having a rant at all, just stating our findings so far.

We do quite a bit of animation right now, and animators are liking it. Our scenes are not that complex, but we do experience some of the crashes when animating that you are talking about, and the other issues as well (except for Digital Assets, which I find good).

However, animators using it have no experience with other softwares (some of them are 2D animators) so they are not aware of the workflows in Maya.

 

The biggest problem for me, when rigging (I'm not generally a rigger) is skin capture and deformation. I guess I'm totally incompetent (quite likely), but I find it hard to get good deformations.

Edited by digitallysane

Share this post


Link to post
Share on other sites

The biggest problem for me, when rigging (I'm not generally a rigger) is skin capture and deformation. I guess I'm totally incompetent (quite likely), but I find it hard to get good deformations.

 

I'm afraid it's not necessarily due to incompetence. 

 

Skinning in Houdini is very much behind current standards. That doesn't mean you can't get good results. But is will take a lot of time, workarounds and custom solutions. For instance I recently needed to rig a simple human mesh exported from makehuman app. I had time so I tried both maya and houdini to get it done. (I only needed skinning as it was used as a bit of a glorified proxy) In maya I had usable result within 15 minutes (their geodesic voxel binding with dual quaternion is really amazing), in houdini I stopped after about 3 hours of painting and fixing, because the default weighting is completely unusable most of the time and tweaking 160 capture regions one by one is beyond my patience. (bones were imported from makehuman and the capture regions houdini created for them were humongous). On top of that, without dual quaternions I wouldn't get the same result without extra bones or deformers anyways (it was quite muscular man so I was losing a lot of volume).

 

Of course it's possible that I'm just incompetent too (even though I have done extensive maya rigging in past and rigged around 10-12 character in houdini over the last 9 months). However if that's the case, then it would only be point against houdini, as finding really skilled people is very hard and if software makes it easy to achieve good results with less experienced, hence cheaper people, me and our producer are all up for it.

 

To wrap it up. I totally agree that getting good (let's not even mention fast) deformation in houdini is currently a bit of an uphill battle.

  • Like 2

Share this post


Link to post
Share on other sites

this surprises me...

I've always been able to get good results using the basic capture methods and point groups to isolate regions...

@Milan - 160 capture regions? for one character?

Share this post


Link to post
Share on other sites

this surprises me...

I've always been able to get good results using the basic capture methods and point groups to isolate regions...

@Milan - 160 capture regions? for one character?

We're using muscles and I find them quite hard to control.

Share this post


Link to post
Share on other sites

ah...muscles...

 

for muscles a really proper workflow has never revealed itself...but the fewer the better...

Share this post


Link to post
Share on other sites
 

this surprises me...

I've always been able to get good results using the basic capture methods and point groups to isolate regions...

@Milan - 160 capture regions? for one character?

 

Well if I exclude facial bones which are not needed in simple cases, then it's around 100 (fingers + toes alone are 60-80 bones, plus at least one twist bone for each limb part). For high end character we'd be looking at much more than that. Even game characters are easily reaching 100 bones nowadays, so it's really not that crazy.

 

Point groups are a must I totally agree with that. 

 

However this is what I was trying to say, number of bones aside. Comparison of identical mesh and skeleton skinned in maya and houdini. No painting no settings tweaks. Just select mesh and bind it. Geodesic Voxel in maya, default settings, Proximity in Hou (I only tweaked drop off a bit). Maya 90% there, Houdini 90% left to do. Capture region are in this case useless, as they never seem to be created at roughly the right size. In this case, they were so small, that they didn't catch any mesh at all apart from fingers. I know this is extreme case and no one expect to get one click final results. Spending 8 hours and 1 hour on cleanup though is a massive difference. Plus I simply can't get proper volume preservation in houdini without extra tricks. 

post-7709-0-60414500-1428343404_thumb.jp

Edited by Milan
  • Like 1

Share this post


Link to post
Share on other sites

pretty sure I've seen in in a Houdini Wishlist thread more than once...

 

but yes - anything that you think should be implemented in Houdini should always be sent in as an RFE - nothing in either forums is actually considered an official RFE, sometimes the developers will implement something or fix a bug that they see in the forums but that isn't standard practice.

 

as for geodesic voxel, I don't see why it can't be implemented fairly easily...

Share this post


Link to post
Share on other sites

Sesi has just fixed the issues with copying animation between rigs if anyone is interested. Can't wait to give it a shot. 

 

 

Aside from that. I was hoping I wouldn't have to say this, but the latest Maya release has (at least in our eyes) effectively cancelled any thoughts for future animating in Houdini. Fully threaded and GPU supported rigs + delta mush support out of the box? oh yes thank you. 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×