SideFX Houdini 15 it out!

[Bullit]

Well Mathaeus criticized issues of Houdini while answering a question i made. It had nothing to do with Maya.
You could criticize the points Mathaeus directly and/or you share the good experiences with Houdini, then us the hesitant ones can choose better. No application is perfect and the weight we make between good and bad, the perceived potential evolution, the trust we have in the vendor is what makes one choose one over another in relation with specific work we pretend to do.

For example you guys have had no issue with constraints that Mathaeus makes reference to? Did you need to bypass the issue? How was it done?

[Mathaeus]

Well, about animation constraints, avoiding gimbal lock and related stuff. As far as I know practices are different, theoretically someone could feel satisfied with Houdini offer, too. However, 3d apps who already have something to show here, let's say Softimage, Maya, Max, Blender, Lightwave, all have very similar solutions, so I think we can talking about standard - which really is not followed by H, especially not when it comes to such important backbone, as constraints are. New animation editor is a nice baby, but, what is driven by H AE, how it's driven, that's another story.
On positive, opposite side, there's Mantra renderer, which occasionally or not, *is* following standards since its PBR times. And, that 'little Arnold' has a much better reputation than H animation part, I think everyone will agree. According to Ubertage video, new Principled shader is 1:1 copy of Disney solution. By the way it's interesting, where things are going when someone really wants success



That's what is displayed when trying to edit some of new shader presets, exactly, one called 'Piano Lacquer'.
Nevertheless, Mantra is even nicer in 15. Finally H is able to subdivide polygon curves (hair and such), so no more nasty workarounds, to get smooth curves in render. Seems to be faster, too. MPlay (let's say render window) displays more of info, and so on.

[Werner]

Hey Werner,

Would you mind sharing your reel or website? I'm curious to see the work done with Houdini on Character Animation!
Question: You guys still have Softimage, but on modeling, are you guys planning on doing it in Houdini or Blender is being added for that?
I know Blender is improving a lot on animation. And while I still like 3ds Max, I'm considering options to not be dependent on Autodesk and Cebas.

Sure. Visit our website or vimeo page
http://www.luma.co.za/
https://vimeo.com/lumastudios/videos

The Cowbell commercial was the last Softimage/Arnold project we did. We are currently working on a new project that will be all done in Houdini, and I will post here once it is done.

Here is a quick modeling test (animated giff) in Houdini. I've set up hotkeys for most modeling tools, and so far it is allot better than previous versions. It needs more work and is far from perfect, but I have confidence that SideFX will get it right.

http://postimg.org/image/meaq2meuv/
http://postimg.org/image/5z66dcpsb/

We hope to do everything in Houdini and render with Mantra.

[Werner]

For example you guys have had no issue with constraints that Mathaeus makes reference to? Did you need to bypass the issue? How was it done?

Maybe the way we build our rigs are allot different than how Mathaeus does it...not sure. I'm on my second rig now, and the animators had some questions here and there (like 'Euler filtering' in Maya or 'Keep rotations continues' in Softimage). That has been taken care of in H15. No problems with constrains...or the lack thereof.

You can either use channel referencing or something called the fetch SOP. I checked the performance monitor and it seems like parenting with the Blend SOP was fastest...so that is what I went with. I'm not saying that these are constrain replacements, but they work for us.

Something that bothers me allot more than constrains is skinning and weight painting (What they call Capturing). It needs to be pulled into the now as it feels old and clunky. There has been discussions around this and I'm looking forward as to what they will come up with to fix it.

cheers,

[dalvirohan]

Hi Werner,

I haven't seen a lot of people do actual character modelling in Houdini but those gifs are really good. I think it's a great way to show that you can use Houdini for all kinds of 3d projects and not just for particles and dynamics, since that's what most people tend to associate Houdini with.

Also the work on your website is absolutely phenomenal.

I think if in the future anyone ever asks me about Houdini I'm just going to point them to your website.

regards
Rohan Dalvi
http://www.rohandalvi.net/

[Werner]

Thanks Rohan,

I'm a huge fan of your work and loved your 'User interface and motion graphics' tutorial. Amazing stuff.

Everything on our website was done in Softimage and Arnold, but that is about to change.

The team spent allot of time playing around with Apprentice, and we are now moving over completely(Keeping Arnold as part of the toolbox moving forward). Give us some time to finish the first project before you
point people to our work.

I am looking for proper screen capturing software and will see if I can show more on the modeling side of Houdini. Can't wait to slap some fur on the character in the animated giffs.

Thanks again for the kind words.

[Werner]

Not Houdini 15, but proof that rigging is possible. Mantragora is a nice guy and very active on Houdini forums. I'm sure he would not mind to jump in and share his thoughts on this subject.

[Mathaeus]

Everything is possible, but for what price....
Just did some really simple tests, using 'simplefemale' from H15 samples, against closest, speed wise comparative SI model I had in my ancient XSI 7.01.
So, these two shows a similar playback of around 50-60 FPS, on my home toy, quadcore i5 machine. While SI model has 1.5 more polygons, about two times more of bones (around 150 against 70 in H, I think), also soo much more complex rig, having a doubled IK/FK chains for arms and legs, around +120 constraints for intermediate bones, so on.
Intentionally reduced both, down to only envelope operators. As it noticeable in screenshot,
H rig includes a cache node (similar to our 'use 3d cache'option, while H has more of nice options here). This adds about 25% to H rig, but only after one playback.
Nice trick btw , but we want it immediately...

Long story short, let's say that Softimage XSI 7.01 shows around... 3 times faster data flow than H 15. Don't have Maya installed, which probably is better for both players, here.

Anyway, not really a surprise. From what I heard, from my personal impressions too, H simply is not focused to real-time performance. If developers of H don't care that much about this part, no one could go so far with individual actions. Simply as that. Also, found H as too much sensitive to changes in hierarchy (this one in H is 'ideal'), or operator structure, while SI simply don't care about these things, in much wider range.
On flip side, H still has a sort of ICE capabilities, while Maya don't - so - nice mix could be Maya for rig and animate, H for all sorts of custom deformations, well if someone is able to build them, or feels any need....
Regarding the price of Maya LT and H import capabilities, I'd say, it's only something related to emotions, fanaticism or what, to make a very very expensive decision, to do not use Maya in mix, these days.

[nodeway]

I'm sure he would not mind to jump in and share his thoughts on this subject.

I'm not a rigger, so nope.

I had also Robocop MK2 Cain rigged. That shit was pain in ass to rig (and model, also in Houdini). But I lost part of the model (upper part of the armor and back) and rig R&D in HDD damage, before I started baking my stuff to the cloud. From rigging only this => http://www.mediafire.com/view/zzzukttpcjvutsx/rig.gif <= and main body hydraulics survived.
I'm still waiting for a better mood to come back and fix it. Maybe some day...

BTW.
Anyone knows how it's with the rights to use something like this in a tutorial? Not give to people full model, but show how to model something like this. Do I have to ask for any approval some company with the rights to the franchise?

[Bullit]

Thanks Werner for the answers.

[Mathaeus]

Anyone knows how it's with the rights to use something like this in a tutorial? Not give to people full model, but show how to model something like this. Do I have to ask for any approval some company with the rights to the franchise?

Ask them, keep the permission on safe place, always. Especially if you're making money from tut.
Once your tut is exposed into wilderness, no way for retro actions. For example from reality, last time someone used my model for such purpose, it was for some Foundry Mari, free tutorial. Had to read 5 pages of contract, sign it and fax it back to them.

[angelous4x]

H15 is a great release. and the guys over a sideFx are pretty awesome. but my issue with Houini is the same.

The UI is just terrible. It needs some serious love,

the interface alone is really (IMHO) the biggest hurtle for most artist and untimely drives them away. doing tut within sops, to do a "move towards Goal" or liner interpolate positions in ice is just a few nodes. In Houdini theirs about (using both sops and pops) its 20+ nodes.

ok but lets look beyond that because you could make assets of the tools you mostly use.

i get a someones scene ( or revisiting a old project) and with in soft, ice i could see the whole tree, and at a glance know where things are and find the all the attribute and what make them up and where they are used or changed.

with in Houdini, its another UI workflow hurtle. you have create attribute nodes out side of the attribute vop, and use the created attribute with in that vop or several vops. and then change that data in another vop sop, then use expressions calling a attribute that changed in the 3rd vop sop in lets say a copy node. or other of nodes. now you have a scene with dozens and dozens of nodes, nodes with in nodes, with expressions in many of them but finding expressions/attributes/ could become a nightmare. But first you have to learn some vex.

where is this expression? where is that attribute?where is that attribute changed or used in what expression in what node?

and whats up with all the crazy icons? (Seinfeld voice) 95% i don't think are needed.

Soft has a better work flow and organization (almost no icons, and when you see it they make sense). the UI is just a lot better everything is done with the least amount of clicks possible. which is why most maya artist could easily move over to soft and start doing more then what they did in maya. but that being said

Houdini is alot faster and stronger then softimage. A Billion f@*kin particles DAMN!!!! Mantra is Superior to Mental ray hands down. and of course the non-destructive work flow is awesome.

but i am still learning Houdini, and this is from a learning/transition point of view. But i am sure if they created a more user friendly UI and allow artist to create more and script/code Nothing (like softimage ice and Fabric engine where you don't need to write KL and use there nodes) they get more artist to jump on board.

i'll shut up now.

[FXDude]

I have to agree with angelous.

One common issue with always being fully procedural, is for example, say a relatively simple asset like perhaps Rohan's rocket is inherited, or went back-to after a bit,

at the scene level you have the rocket, the hills, a light, a camera, etc...



but then going in to edit the rocket is like.. okay... this connects to there and connects to there to make this modified by that expression... and any given 'element' down to what defines it, can only be something quite abstract within a given graph.



Which often needs to be (re-)deciphered to go in and change things, or to figure-out where different things start and where they end, without immediately convenient ways to just start from a given point, or even when doing so, without everything becoming frozen (no history at all) and unavoidably then having dependancies on external files. Like always dealing with almost completely everything, or almost nothing at all.


So while that level of proceduralism can only be invaluable for a great deal of things (arguably more than ICE), it can also quickly become quite a bit to manage.

Which may be why searching for "Houdini" in Vimeo consitently brings-up page after page of the same type of things you would typically find when adding "ICE" after "Softimage" in the search field.

In Rohans (BTW very well explained) Clip about proceduralsim, (with the table modeling example) he goes on to say how Houdini -can- be slower for certain things (not refferring to FPS) because you have to think about how to go about building things, also mentionning how it then allows for the benefit of having everything ... procedural.

Benefits which can only be obvious, not unlike what was the entire point of ICE (or the comparatively simple construction stack).

But sometimes (if not most of the time) you just want a table (and things), and translate up the top faces if you ever happen to want it higher, while keeping the more intricate procedural effects for the normally very few more intricate things.

So maybe indeed as Matheus suggests, that some sort of Maya/Houdini combo (while forgetting about c4d friendlyness) might come closest to what SI was all about.. but (sigh).


PS: Mantagora, I mentionned your excellent work, but it is really quite excellent! Kudos!

[Mathaeus]

So maybe indeed as Matheus suggests, that some sort of Maya/Houdini combo (while forgetting about c4d friendlyness) might come closest to what SI was all about.. but (sigh).

I was talking exclusively about playback and characters ..... For scene building in H, tried several variances for this thing, here on forum, finally ended with many small scenes loaded into one, around five H bgeo files and dozen of objects from SI. Imho main comparative advantage of H is '3d compositing' ability, to load 3d objects, in same way as 2d comp app is doing with images. Doesn't matter from where objects are coming, when update is only a new export in another app, reload in H.
This approach, btw, occasionally fits to descriptions in Jordi Barres PDFs. Also it fits to my own 'theory' about 'stepped proceduralism' . Basically it says that forms, where method of creation has some usable effect later, actually are rare forms (particles, trees, hair, staircases and such), also methods are already implemented in almost all 3d apps, most of time more efficient than in H.
On opposite side there are 'steps' where complete element has to meet only some criteria, to be able to be processed later, while next step is able to adapt to previous one. Like mesh of character, getting an envelope - which was, imho, first serious break of old fashion 'keep it procedural from bottom to top', naive idea. People will remember methods from 90's, to utilize NURBS blends for animated characters, or to use meta balls. Both of is 'fully procedural' - but never worked enough good for entire process.

[FXDude]

For scene building in H, tried several variances for this thing, here on forum

Also pretty cool

[Mathaeus]

A Billion f@*kin particles DAMN!!!! Mantra is Superior to Mental ray hands down. and of course the non-destructive work flow is awesome.

You'll want a really really strong machine for that. Found Houdini and Mantra a lot more memory hungry than SI and MR, or SI and 3delight - don't know about Arnold. For pic linked above, had to upgrade my RAM from 8 to 16 gigs, just to be able to render. Had hard time, tryng to get enough fast displacement on rocks, finally rendered without any displacement. Yeah I know these numbers are joke, but, 8 gigs and SI were enough through the years, for all sort of playing with ICE, millions of ICE strands, what not.
Heard it's better with memory on Linux, but don't have plans to switch to Linux...
Did some displacement test in H 15 (which took a good time in v14), so a lot faster pre - processing time, now, but still very memory hungry.
By the way, Mantra has specific way of working, it's separate process of H. That means, once Mantra started to render the frame, you can proceed to work in H on something else, but, entire transferring process is slower. From what I read, only Katana has this solution, too. Something probably nice for long time renders and big studios.