aRtye

amm
That Maya speed is not only for super complex rigs for movies - it allows to animate at double (or more) FPS, getting a much smoother movement, it allows to have a multiple transformation structures always available for animating, so on.

Most studios have custom solutions as parallel rig evalution isn't robust.

Here Maya Parallel works really nice and predictable, much smoother in Maya 2017 than in 2016. Perhaps a part of that ‘predictable’ belongs to around ten years of my playing with lazy evaluation system in Softimage. Yeah, some experience is needed to recognize is it just a ‘too lazy’ evaluation sometimes, or there's some error on road. Such system is always sensitive to hierarchy structure.
Really don't know what happened with these studios, which practices they had before and how that fits to current Mayas, is there something else that counts. Not enough info for anything useful for me (as usually).
Let's say from Softimage perspective, reliability of Maya Parallel against SI (classic) kinematics is comparable to ICE - Houdini SOP-VOPs ratio. Some parts are different but generally, after some time it's possible to be focused on creation, having a ‘rules’ stored somewhere in ‘intuitive part of brain’.

amm
Some parts are different but generally, after some time it's possible to be focused on creation, having a ‘rules’ stored somewhere in ‘intuitive part of brain’.

Yes of course - that's the old system 1 / 2 at work.

Overall if Maya is fantastic that is great, if Fabric is great that is cool but it would be silly to try to diminish what Houdini is trying to achieve, so any constructive comparisons will help the devs make the ‘booleanSOP’ of animation.

This makes me wonder how many people here has actually built a full production rig and animated it in Houdini.

We do this almost on a daily basis, and we are extremely happy using Houdini for this. The incredible non destructive workflow is second to none. The flexibility of changes during the whole process is something Maya users can only dream of. Biharmonic capturing is a massive improvement and brings down weight painting allot. Adding secondary deformation
at any stage is amazing….without having to scrap skinning or rebuilding parts of the rig.

We have another division at Luma that uses Maya mainly, and they are currently going through the struggles of learning Houdini. I find this very enlightening, because it gives us fresh ideas/RFE's to submit. This will help SideFX improve Houdini.

Like all other software Houdini is not perfect, but over the last two versions, SideFX has really worked hard to improve this. Ask anyone in Luma Animation if they want to move a project over to Softimage/Maya or do it in Houdini, and you will be pleasantly surprised. This was not the case six months into learning Houdini, but took time…like all areas of Houdini. Yes, workflow can improve overall, and submitting RFE's is almost a weekly task for me.

Again, rigging is not perfect, but in my mind, it is no longer the problem area in Houdini.

Building HDA's for the rigs, is where the problem comes in.
It takes way too much time and effort to build the interface for animation.
On a tight deadline working on commercials, this is always the hassle.
Overlapping selection sets that is portable would be god sent to improve setting up HDA interfaces. This area can improve greatly…not just for rigging but HDA's in general. Remember how easy building and saving Softimage compounds was. This is not the case in Houdini.

Animation itself seems to be less of a problem as it is very similar to other applications now days. Again, not perfect, but so was Softimage/Maya.

My observance is that people who had no problem learning various applications during their career has less of a problem than young newcomers who studied Maya at collage. The find it hard to let go of the only workflow they are comfortable with. Positive attitude goes a long way.

Take time to learn the tools and then you will understand the immense power Houdini holds. After 16 years in the 3d industry, I am happy to go to work again.

http://www.luma.co.za/ [www.luma.co.za]

Sigh. So this is the nth thread on the internets, where people critique without contributing with something substantive and others praising the tool.
I thought I'll learn something from this thread and see some good ideas popping, like for example when I investigated the matter of creating a corrective shape I have to use a command line (which is not the bogyman, but it's not the most modern approach either when it comes to rigging/animation), or how we'd benefit from a shape manager similar to other apps (or a better one if you can come up with a better one) or maybe we wouldn't, either way, a discussion about concrete problems and solutions.
Alas, it's a water-cooler chat. It could be fun to read but its usefulness stops there.
And to live by example, I'll bring up the aforementioned issue and I hope to see proposals for improvements or other existent solutions if there are any to the corrective shape issue.

I personally can't understand how anyone can complain about rigging in Houdini specially compared to Maya, where rigs are so much harder to build and maintain. I feel like most avid maya riggers simply love being miserable secret order of pymel monks. They clearly control world around them this way.

I wouldn't focus too much on Maya's parallel speed also, not because speed is not important, but because it's a new thing, which many people don't even use yet, so if Maya has built its reputation among riggers and animators, it was based on something else. State-of-the-art has to be matched one day if SESI wants to compete, plain and simple. Then some more esoteric issues like animation workflow or advanced rigging techniques (constrains) are much harder to nail in design.

Animators are photoshop people. They like smooth, not intrusive experience. Houdini's architecture is currently not good with that. Even I feel like animating objects in Maya is easier, despite the fact I almost never use it, while Houdini is pretty much every day on my screen for 12 years.

We, as many others I'm aware of, genuially animate in Maya for a number of reasons from which only few are technical. Our riggers would love to work in Houdini, they've picked some bit of it recently, and constantly begging about Houdini engine in Maya for animators. Our animation lead has built his own muscle system, which is currently in production. He's made VEX based rigs, and solved problems he couldn't even imagine to solve in Maya. Still at the end of the day, back to reality, and he fixes Maya rigs.

So in that landscape we do animate from time to time in Houdini. The animator seats to, like just recently, camera layout. He spreads Houdini into two screens, he grabs camera handle, handle moves, camera stays, he touches handle again, misses it, accidentally selects another object, Houdini jumps into another subnet, curves disappear from animation editor, he's lost. Oh, I have to pin pane and curves! He does that. Now camera moves but gizmo stills. Oh, constrains! They break gizmo's display. Ok, somehow he manages it. Preparing your UI for animation is intellectually challenging, he has to choose which pane to link and pin and whatever, just to make it stable while he jumps between objects and handles. Then Houdini's infamous bug of ghost meshes starts to magically appear, he's confused which object is which, where is his camera among them, there are tree different camera gizmos among whom none of has correctly scaled frustum which doesn't help either (and two are hardly usable in current state). Also, one animation object doesn't update at all apparently because animation editor is on another screen without window focus. Once it happens, he has to restart Houdini to make it work again. This is happening in 600 something build of Houdini 16, after two years of its development.

Did I fire bug report? No, because cases are hard to reproduce, we animate in Houdini rarely, I have enough problems in different departments, and I wouldn't even notice these bugs myself, if I wasn't seating behind this poor guy. And I read Mark Alexander many posts explaining how viewport update policy is extremely complicated due to different contexts asking for different displays anyway, so I understand it's not a problem of bugs, but architecture working against constant stream of interaction from the user on different Houdini's levels.

Houdini flexibility requires sophistication, but sophistication beats interactivity. Interestingly enough this stays true for pretty much any node based system to some extend. You want friendly user interaction, you need to put a proxy between the user and your objects' model. Isn't it what Maya Gui is?

Say, new modeling tools in SOPs, less procedural and flexible, but somewhat more interactive, aren't they?

To put it shortly, we need flexibility in rigging, but plain simplicity in interaction. This is what new animation software are doing these days. Houdini for animators should look and feel like iapp. Not sure if SESI self esteem can handle that.

symek
I wouldn't focus too much on Maya's parallel speed also, not because speed is not important, but because it's a new thing, which many people don't even use yet, so if Maya has built its reputation among riggers and animators, it was based on something else.

Well, Maya always been fastest option available to buy by anyone, of course when it comes to evaluation of rigs and deformations. Parallel thing is just a last feature.

In the end the question boils down to if you can make money off the animation tools in Houdini in the commercial world. Performance and functionality problems ultimately become cost issues, so at this stage are we sure that SESI are committed to being fully competitive with industry leaders Maya et al? Im yet to hear CB commit to it like the modelling tools have been. Perhaps it's a question of animators not being vocal enough on the forums Modellers rule the roost, animators are weak!

@symek the SopSolver is usually responsible for the frozen meshes in the viewport. Wondering if that's the case for the animators too.

“In the end the question boils down to if you can make money off the animation tools in Houdini in the commercial world.”

Kind of this.
It should be possible to have animation system such as you can get in couple animators do some basic training in a week or two and have them animating in no time. If they need a lot longer course to get into speed it will be hard to get more animation transitioned there…
And lack of riggers out there isn't helping at all.

aRtye
so at this stage are we sure that SESI are committed to being fully competitive with industry leaders Maya et al?

The natural segue from this snippet of thought would be the eternal egg & chicken dilemma: do your clients make your product or does your product make your clients?
I'd argue that in a market in which you're both feet in, you should not wait for clients to make your product, especially when they have alternatives. If your resources allow it. No one buys a s/w product hoping it'll get better. They buy it for what they need the most and hope it gets better in other areas, but to get new clients you have to improve your weak areas because you already have clients for the strong points. Waiting your existent clients to become interested in the areas you're weak at is a risky play. Remember, they've been having alternatives for a long time had they been interested in areas you're weak at.

And as far as SideFX is concerned, a mere half-educated guess, they'd probably spread too thin, so they have to tackle different areas in turn.
Or maybe it's a scalability issue and one can't grow a 3d app in all areas in the same time regardless of how many resources are available. I don't know… SESI could answer this one probably.

I'm going to speculate that Sops 2.0 aka compiled surface operators (Tops perhaps) is required to be get evaluation speed up and that Chops needs to be more multithreaded as the last attempt of DeformSOP OpenCL was a non-starter IIRC.

Therefore, with hindsight of how the modelling tools are going through their rewrite, we are looking at 2-4 years to get Houdini at a competitive level. That's 2020.

Also benefiting from hindsight, I have to say your estimation is too optimistic, assuming these two matters are of equal complexity and development efforts will be constant. In my appreciation Houdini's non-procedural modeling is nowhere near where it should be as far as I'm concerned. Hopefully version 2017 will bring more, apart from the updated topobuild which, although nothing to scoff at, it's not really what I'm having in mind here, else we're looking at ~2020 ETA, back from 2014, so a 6 yrs journey.

edit: Forgot to add that thus far SESI's philosophy seems to be “only current clients decide what the product is going to look like” as it seems that only paying clients are allowed into alpha/beta testing. As a mere potential client, all I have left is hope - the hope the development goes into a direction I like. The only reason I'm still hanging around is the potential I see in H, not current capability. Current capability has trapped in almost no “traditional artist” so far. True, I'm no “pro” modeler, but do you think if tomorrow I'd be hired as a 3d modeler, which would make me a “pro” by definition, I'd use Houdini for anything? So yeah, it all boils down to whether you can make money with a tool or not.

Symek could you please post this to SESI team as animation RFE?

We got very similar interaction from animators. They were surprised that they cannot move multiple objects at once in the year 2017, but compared to overall interaction workflow this is a pain. They were used to XSI, then had to downgrade to Maya. Now they are refusing to downgrade to Houdini interaction experience anymore.



Exactly this:

symek
So in that landscape we do animate from time to time in Houdini. The animator seats to, like just recently, camera layout. He spreads Houdini into two screens, he grabs camera handle, handle moves, camera stays, he touches handle again, misses it, accidentally selects another object, Houdini jumps into another subnet, curves disappear from animation editor, he's lost. Oh, I have to pin pane and curves! He does that. Now camera moves but gizmo stills. Oh, constrains! They break gizmo's display. Ok, somehow he manages it. Preparing your UI for animation is intellectually challenging, he has to choose which pane to link and pin and whatever, just to make it stable while he jumps between objects and handles. Then Houdini's infamous bug of ghost meshes starts to magically appear, he's confused which object is which, where is his camera among them, there are tree different camera gizmos among whom none of has correctly scaled frustum which doesn't help either (and two are hardly usable in current state). Also, one animation object doesn't update at all apparently because animation editor is on another screen without window focus. Once it happens, he has to restart Houdini to make it work again. This is happening in 600 something build of Houdini 16, after two years of its development.

aRtye
I'm going to speculate that Sops 2.0 aka compiled surface operators (Tops perhaps) is required to be get evaluation speed up and that Chops needs to be more multithreaded as the last attempt of DeformSOP OpenCL was a non-starter IIRC.

Therefore, with hindsight of how the modelling tools are going through their rewrite, we are looking at 2-4 years to get Houdini at a competitive level. That's 2020.

My impression is, that there is something different in evaluation of transforms, compared to SI or Maya, also in the way how transforms are transferred to SOP network. H is surprisingly sensitive to levels of hierarchy. Also, hierarchy in H is surprisingly immediately responsive, which is (ironically) bad sign, of not everywhere present lazy evaluation, like in mentioned two. In Maya, ‘parallel’ is tied to transforms, second component is open cl evaluation of deformations, last component is ‘no write back’, evaluated deformation is just displayed by Viewport 2. Of course such system has a lot of drawbacks: only exact number of seven built in deformers is supported (where two of them usually are not considered as deformers). To keep the ‘no write back’, any not-supported deformer ‘on top’ is forbidden. On other side such configuration matches (more or less) to evaluation of game engines, so Maya (again ironically) behaves as ‘cockroach able to survive the high level of radiation’ emitted by two or three dominant game engines, today.
Also, Maya's ‘totally flat context’ in Houdini words, while it's probably base for high speed evaluation, is also a base for traditional horrible moment when it comes to managing these networks, while Houdini system of separated networks is clean and pleasant, compared to anything in Maya.

Long story short, it seems that nobody could expect to compete in everything against anyone. There are somehow immanent good and bad sides.

Imho, software like Houdini should emphasize Houdini good points, differences against rest of 3d world.

And yeah there is that programmer vs artist cultural difference, well known for artists who worked in small gamedev teams or they have programmer in family (I'm ‘covering’ both ). What's irrelevant for one side, it is highly valued for another, so on. While I'm on artist side, I'd still like to see H as mainly ‘programmers corner’, that's simply how my perception is configured, from long time ago.

amm
Imho, software like Houdini should emphasize Houdini good points, differences against rest of 3d world.

I salute your divergent opinion, but by looking around you should know that it's not going to happen. SideFX doesn't seem to see it this way, or they'd be pouring all their dev. on simulations.
Every 3d software I know tends to migrate to the position of “pipeline hub”, even s/w that started as highly specialized, like Modo, started adding rigging & animation and even sims. Hell, even Photoshop got into 3d painting.

Houdini will not and cannot remain good at only the things it's already good at if it wants to grow significantly.

McNistor
Houdini will not and cannot remain good at only the things it's already good at if it wants to grow significantly.

Houdini is at a sweet spot right now, kind of just needs more refinement, optimisation and rigging tutorials

I don't know about a sweet spot. Non-procedural modeling is still really weak, it lacks lots of tools and a robust transformations sys. I'm assuming, or maybe I shouldn't assume anything since SESI themselves said they'll make Sub-D modeling a top-notch solution, it's traditional modeling is on a to-do list.
Procedural modeling is of course strong.
Rigging is strong but indeed needs performance enhancements. And beside that, there's a ton of room for innovation here - say better muscles, ie more anatomically correct in shape and behavior.
Animation toolset has lots of room for improvement - even compared to existing or dead programs (XSI comes to mind)
Simulation is strong obviously, but even here there will always be a desire for more speed and accuracy. And cloth sim.
Naturally, it's a matter of perspective, but I for one think it's just shy of reaching that sweet spot. Maybe, hopefully v.2017 will brush with it.

The sweet spot is defined if you can make money off the tool and it is not the tool that is hindering work-flow. If I were you I'd be using all the other tools and not waiting for Houdini, it'll most likely never all those things you are seeking.

I assumed that, and still hold that belief, all the noise starting with H14 was around bringing additional value to Houdini, other than what “veteran” Houdini users have used the tool for traditionally.
And I don't wait for Houdini, I do exactly what you'd be doing if you were me. I'm around here (clearly I have to explain this) to contribute with my perspective and ideas in the context described above. Some of them have even been implemented and if I'm wasting my time contributing to a tool I might never use, that's fine, I don't hold anyone responsible for my choices. Cheers!

btw: I'm assuming you've been/are in beta and I'm interpreting what you're saying as you pointing to an invisible writing on the wall regarding v.2016.5/2017

Mirko Jankovic
I was wondering if there could be some insight on this matter.
How's the current state of character animation in Houdini.

“Evolving”, that's the current state.
Thing is, I don't think it's realistic to expect Houdini to just drop on the scene with everything fully ready for character animation.
If there's interest, it will come. They (SESI) are doing a lot of effort, some stellar, some hit-and-miss (the new constraints, amazing in features but lacking in design, IMO). And the only way to improve on this is people using it, getting involved. If SESI are not making money from people doing animation, then at some moment they might stop trying to sell to them.
Studios that are willing to use Houdini because of its amazing flexibility and workflow features, for which they are willing to accept that some of the character features are a bit of WIP (still, the character and rigging toolset is solid at the moment).
Like we do, for example (we use Houdini exclusively for character animation, but then our projects and pipeline are not necessarily the typical ones), or like @Werner does at Luma.
A good point about this is that, if someone gets involved at this moment, they have a chance to actually help shape the tools and workflows, which is something unthinkable when it comes to other packages.

“And the only way to improve on this is people using it, getting involved”

This kinda is exactly what I had in mind.
If it is possible to jump into it, have some nice rigs and just start animating and using it, it would be a start at least. Checking out animation tools and what is missing, giving some feedback hoping to get them more polished (and hopefully close to Softimage level)
But as I'm really honestly not too much interested into going into rigging first and then make my own rigs, too busy and too tired for learning more rigging in another software… lack of nice proper rigs for play with is a bit of issue and the one hopefully could be addressed.
Maybe create donation library somewhere and anyone that can and is willing can share some nice rigs.
Then that would at least give possibilities to start using them around, maybe for like 11secondclub and similar playgrounds and would give them proper testing as well. Not rigs only but animation workflow as well.
But as mentioned before at the moment it is a bit harder to just grab and start animating.