Article by SESI's Tony Cristiano
http://vfxworld.com/?atype=articles&id=2496 [vfxworld.com]
requires free registration
The Great Divide in CG Facilities: Adapt or Die
27713 23 7- goldfarb
- Staff
- 3459 posts
- Joined: July 2005
- Offline
- markholloway
- Member
- 2 posts
- Joined: July 2005
- Offline
This article about slashing budgets, salaraies, and re-using digital assets is interesting considering it's written by Side FX's COO, one of the most expensive 3D software company out there. Maybe I'm missing something? But what I'm getting out of this is that I should be using Lightwave 3D as my animation package instead of Houdini, among other things.
- goldfarb
- Staff
- 3459 posts
- Joined: July 2005
- Offline
- edward
- Member
- 7868 posts
- Joined: July 2005
- Offline
I'm biased but I think it's a misconception that Houdini is the most expensive. For some reason people forget that Houdini (Master) includes a compositor as well. Plus it can inter-operate with Houdini Escape which is far cheaper. This means that for a studio of 30 or more people with an appropriate high ratio of Escape/Halo to Master seats, the unit cost per person is now much lower, even competitive to some other packages.
- Simon
- Member
- 2199 posts
- Joined: July 2005
- Online
The way I see it everytime I make a digital asset rather than buy a plugin we make some of that money back. Plus, as arctor says people are way more expensive than software, just use the best software for the job. That may often be Lightwave but if it's Houdini and it's for a good reason (and OTL's are a very good reason) then the cost balances out.
The trick is finding just the right hammer for every screw
- el_diablo
- Member
- 133 posts
- Joined: July 2005
- Offline
Well, I really cant understand how Houdini Master can be deemed expensive as its still covered financialy by doing one vfx project in US. In my realm (Central-europe, Croatia) its extremly expensive as vfx projects here range from 500-5000$ depending on complexity. There is higher budget work, but its far and between. So ideally I would need full 6 months of work just for the license which is completly unrealistic. But I'll save my pennies eventualy .
BTW. Did SideFX sell any licenses at all to Croatia? I would think not, so my company might be the first! lol :wink:
BTW. Did SideFX sell any licenses at all to Croatia? I would think not, so my company might be the first! lol :wink:
- MatrixNAN
- Member
- 405 posts
- Joined: July 2005
- Offline
Hey Guys,
Well I would not begin to compare Houdini and Lightwave to each other. Houdini is so much more powerful for VFX and Lightwave ever dreamed of you can not put them in the same realm. I own a full commerical copy of Lightwave 8 and I have used it commerically on some projects but I have gone over to XSI, and I am saving up money for Houdini. Lightwave does not even have a node based shading system which is really important for VFX. Also their particles are very primative at best and controlling them to any detail would require large amounts of L-scripting which would cost you more money than what it costs to buy a copy of Houdini. If you are talking Character Animation then XSI, Houdini, or Maya are better because the rigging system in Lightwave is so bad you have to buy a plugin just to actually get it to work right. Basically every package but Houdini is going to require lots of scripting for their digital assets. This means alot more people which with atot more computers and alot more software licenses. Yes you will have to pay more for a single Houdini Artist but that artist alot of times does the work of several on a production for creating custom tools and getting around the hard problems without spending alot of money to develope a custom in house tool system from scratch.
Cheers,
Nate Nesler
Well I would not begin to compare Houdini and Lightwave to each other. Houdini is so much more powerful for VFX and Lightwave ever dreamed of you can not put them in the same realm. I own a full commerical copy of Lightwave 8 and I have used it commerically on some projects but I have gone over to XSI, and I am saving up money for Houdini. Lightwave does not even have a node based shading system which is really important for VFX. Also their particles are very primative at best and controlling them to any detail would require large amounts of L-scripting which would cost you more money than what it costs to buy a copy of Houdini. If you are talking Character Animation then XSI, Houdini, or Maya are better because the rigging system in Lightwave is so bad you have to buy a plugin just to actually get it to work right. Basically every package but Houdini is going to require lots of scripting for their digital assets. This means alot more people which with atot more computers and alot more software licenses. Yes you will have to pay more for a single Houdini Artist but that artist alot of times does the work of several on a production for creating custom tools and getting around the hard problems without spending alot of money to develope a custom in house tool system from scratch.
Cheers,
Nate Nesler
- talos72
- Member
- 112 posts
- Joined: July 2005
- Offline
I do find it interesting that more and more software developers are trying to implement proceduralism and node based tools (usually as an after thought on top of an architecture that doesn't really support it) while Houdini has always been a procedural system from ground up. In some ways, other software companies are playing catchup.
Yes, using and creating digital assets can save your butt during tight production schedules and Houdini is ahead of the game in that department with OTLs. That's simply the truth.
I take using nodes to creates assets over scripting everything if I can help it. 8)
Yes, using and creating digital assets can save your butt during tight production schedules and Houdini is ahead of the game in that department with OTLs. That's simply the truth.
I take using nodes to creates assets over scripting everything if I can help it. 8)
- peliosis
- Member
- 175 posts
- Joined: July 2005
- Offline
Yes, and all the people talking about how hard and technical houdini is…
some time ago I was interested in maya and asked my maya friends many questions.
The main answer for “how to?” question was something like “oh, thats very simple, just about xxx lines of code”
That's funny, I don't know houdini even good, and I'm able to create in minutes things I thought about for days using different software.
I spoted a nice link at xsibase today, about rig scripting :
http://www.xsi-blog.com/?p=29#more-29 [xsi-blog.com]
cheers
Peter
some time ago I was interested in maya and asked my maya friends many questions.
The main answer for “how to?” question was something like “oh, thats very simple, just about xxx lines of code”
That's funny, I don't know houdini even good, and I'm able to create in minutes things I thought about for days using different software.
I spoted a nice link at xsibase today, about rig scripting :
http://www.xsi-blog.com/?p=29#more-29 [xsi-blog.com]
cheers
Peter
- jjstanley
- Member
- 124 posts
- Joined: July 2005
- Offline
One of the thinks that Chris kept repeating in the article is effeciency and that is were Houdini really shines. The cost up front is slightly more although as pointed out earlier when you add up everything it gives you it really is not bad. When it is time for production Houdini is far ahead of any of the competion from my experience. In the work place time is money and although you can produce nearly any effect with any decent software Houdini will do it faster and cleaner. That saves a lot more money than the initial cost difference.
- some
- Member
- 15 posts
- Joined: July 2005
- Offline
While Houdini is one of the best frontends to PRMan and seems that the OTLs are working fine, few other very important things are too unefficient ( at least for my taste ).
The overal modeling workflow is unbelieveble heavy - i cant see how people will choose Houdini for organic modeling over Silo, XSI or Modo. Even Maya with its relativelly weak polygonal toolset is way mutch better option. Laying down tons of SOPs to get a simple result is not an option at all.
C'mon, the modeling is all because of the artistic side of the things and the modelers are artistic people mostly. They are screwed-up by any technical stuff.
The overall viewport behaviour is another negative reason - extremelly slow subdivisions, clumsy manipulation tools and non customizable navigation hotkeys and mouse buttons.
Houdini has some nice SOPs doing cool things and sometimes modelers ( or FX guys ) are using them, but thats not enough at all.
The rigging is another weak area. H7 put some fresh air in the room, but the rigging is still weak. Just how the transformation system works in Houdini is not good for rigging purposes. Maya is totally unbeatable here - the transformation system and the overall hierarchy are mutch better designed there. Also the deformers in Houdini need a lot more work. Maybe you will disagree with me here, but i think that those of you, hard involved in the character setup can understand my point.
I can see how SESI is picking good features from Maya ( especially in H8 ) and thats all good, but i think we need more and need it soon - Maya7 is coming and this will change a lot some things.
Animation. The animation must be easy - it must be the easiest and most strightforward thing in the software package. Most of the animators are very weak technicians, some of them have not even basic understanding about 3D space, etc. ( and i think it should be in this way - they, the animators, must care about the animation, not about the tools ).
Houdini must come with well automated animation tools right out of the box ! Cant really see how the overall CHOPs workflow is going to match what the good animation workflow consists of.
Dynamics - the particle system itself is just great, but outside of that Houdini has almost nothing to offer. We must cheat at any step. Is this efficient ?
But yeah … i know, i know - the DOPs are coming.
Customizability - all the UI should be totally scriptable, every single tool, node and action also. Powerfull scripting language - thats the most powerfull tool one application can offer - we can program everything inside the software to work for us in the way we like it. Thats really efficient.
Rendering - while Mantra has bunch of nice features, its motion blur still totally sucks, … why ?
Based on what i said above and regarding the point about efficiency, i will ask - which is better ?
To use other applications, wich offer better content creation tools, or to use Houdini and to compensate with good pipeline workflow ?
Or maybe to use both of them ?
From what i can see around me i can say that pipeline based on multiple animation platforms is not really efficient. Too mutch extra work and restrictions are going for and back between the packages - not good.
Thanks for the attention.
P.S. Keeping in mind all the strong points of Houdini, i'm talking here only about what i dont like and want to be improved.
The overal modeling workflow is unbelieveble heavy - i cant see how people will choose Houdini for organic modeling over Silo, XSI or Modo. Even Maya with its relativelly weak polygonal toolset is way mutch better option. Laying down tons of SOPs to get a simple result is not an option at all.
C'mon, the modeling is all because of the artistic side of the things and the modelers are artistic people mostly. They are screwed-up by any technical stuff.
The overall viewport behaviour is another negative reason - extremelly slow subdivisions, clumsy manipulation tools and non customizable navigation hotkeys and mouse buttons.
Houdini has some nice SOPs doing cool things and sometimes modelers ( or FX guys ) are using them, but thats not enough at all.
The rigging is another weak area. H7 put some fresh air in the room, but the rigging is still weak. Just how the transformation system works in Houdini is not good for rigging purposes. Maya is totally unbeatable here - the transformation system and the overall hierarchy are mutch better designed there. Also the deformers in Houdini need a lot more work. Maybe you will disagree with me here, but i think that those of you, hard involved in the character setup can understand my point.
I can see how SESI is picking good features from Maya ( especially in H8 ) and thats all good, but i think we need more and need it soon - Maya7 is coming and this will change a lot some things.
Animation. The animation must be easy - it must be the easiest and most strightforward thing in the software package. Most of the animators are very weak technicians, some of them have not even basic understanding about 3D space, etc. ( and i think it should be in this way - they, the animators, must care about the animation, not about the tools ).
Houdini must come with well automated animation tools right out of the box ! Cant really see how the overall CHOPs workflow is going to match what the good animation workflow consists of.
Dynamics - the particle system itself is just great, but outside of that Houdini has almost nothing to offer. We must cheat at any step. Is this efficient ?
But yeah … i know, i know - the DOPs are coming.
Customizability - all the UI should be totally scriptable, every single tool, node and action also. Powerfull scripting language - thats the most powerfull tool one application can offer - we can program everything inside the software to work for us in the way we like it. Thats really efficient.
Rendering - while Mantra has bunch of nice features, its motion blur still totally sucks, … why ?
Based on what i said above and regarding the point about efficiency, i will ask - which is better ?
To use other applications, wich offer better content creation tools, or to use Houdini and to compensate with good pipeline workflow ?
Or maybe to use both of them ?
From what i can see around me i can say that pipeline based on multiple animation platforms is not really efficient. Too mutch extra work and restrictions are going for and back between the packages - not good.
Thanks for the attention.
P.S. Keeping in mind all the strong points of Houdini, i'm talking here only about what i dont like and want to be improved.
- edward
- Member
- 7868 posts
- Joined: July 2005
- Offline
some
The rigging is another weak area. H7 put some fresh air in the room, but the rigging is still weak. Just how the transformation system works in Houdini is not good for rigging purposes. Maya is totally unbeatable here - the transformation system and the overall hierarchy are mutch better designed there. Also the deformers in Houdini need a lot more work. Maybe you will disagree with me here, but i think that those of you, hard involved in the character setup can understand my point.
I can see how SESI is picking good features from Maya ( especially in H8 ) and thats all good, but i think we need more and need it soon - Maya7 is coming and this will change a lot some things.
Could you expand on how the transformation is not good for rigging purposes? The only difference I see wrt Maya is that one is bones and the other is joints. With bones, the axes is clearly defined and is usually much close to what a rigger wants right after creating their skeleton.
When you say “deformers”, which deformers are you actually referring to? The actual bone deformer?
Which good rigging features are you referring to in H8? You mean H7?
In terms of character, weighting in Houdini seems to be much cleaner. From the riggers I've talked to, weighting is still the biggest bottleneck in the workflow. In Houdini, one can transfer weights easily in light of topology changes in the skin.
I think one also has to consider the advantages of digital assets.
some
Animation. The animation must be easy - it must be the easiest and most strightforward thing in the software package. Most of the animators are very weak technicians, some of them have not even basic understanding about 3D space, etc. ( and i think it should be in this way - they, the animators, must care about the animation, not about the tools ).
Houdini must come with well automated animation tools right out of the box ! Cant really see how the overall CHOPs workflow is going to match what the good animation workflow consists of.
Digital assets greatly play into this as they just need to know how to pose (ie. Pose tool) a character, the controls on their character asset, set keyframes, and use the channel editor. Plus in H7, there is takes for one to explore layers of keyframes. Mplay blocking in H7 also greatly simplifies the animation blocking process.
CHOPs really don't figure into the bread and butter animation process at all.
- some
- Member
- 15 posts
- Joined: July 2005
- Offline
- edward
- Member
- 7868 posts
- Joined: July 2005
- Offline
- some
- Member
- 15 posts
- Joined: July 2005
- Offline
- MatrixNAN
- Member
- 405 posts
- Joined: July 2005
- Offline
Hey Guys,
Well honestly I hate Maya's rigging so to me that would be really bad thing to copy Maya. I think Maya's character system was out dated and was a good reason for them buying Motion Builder. XSI has a better Rigging and Character Animation System than Maya. I love Houdini's Rigging system over Maya. The reason why is because I can make a complex rig in Houdini without having to do alot of scripting and having to write alot of Maya SDK. Also they way I can build node networks and turn them into OTL's gives me a ton of flexibility with their rigging. Maya breaks down at the node network when you go to duplicate it because rarely maya duplicates the actual nodes and you get something different so that means you have to then write a custom MEL script to go through the steps to recreate that node network. I also ran into something fun in Maya with rigging and that was how many relationships it could handle at once on a character rig with the skin binded in Maya 4.5 they had the nasty problem of if you added the 5th relationship the entire character and all of its parts exploded across the screen and there was no undo because undo was frozen. So you just had to load back up your last save. Also in 4.5 they had the problem of twisting the skeleton through itself everytime you loaded the file. Which of course was no way like you saved it and would then throw off all of your expressions with your custom made graphic icons. So then you would have to go back and redo all of your expressions everytime you saved. In houdini I can create as many relationship as I want and Houdini has never blown apart on me. The only thing I have ever have to worry about is running out of ram or nodes corrupting. Locking nodes seems to keep this from happening but it does tend to happen from time to time. I am not really sure why.
I would also disagree with you on the modeling. You don't model in the tradional since in Houdini like the other packages you listed except for XSI. XSI also bogs down if you do not freeze its node chains. aka delete history in maya or lock nodes in houdini then select all inputs and delete the nodes. In houdini I model completely differently with a totally different workflow then I do in Silo, Modo, XSI, ZBrush, Blender, Lightwave, Maya, etc, etc, etc. Houdini can be your fastest modeling tool once you setup your own system of OTLs. You can use all aspects of the application to model with. I do have to say the fact that you can't go to quads but you do triangles and the fact their is not a quad clean up tool is a problem and does kind of throw the money wrench into the gears so to speak. Also the no displacement shader baking on Houdini Polys is a problem too.
I don't really think the Animation side is a problem because you don't have your animators that don't know anything about 3D actually get under the hood. You have TDs, Riggors, Smart Technical and Artistic persons to it with OTLs and Node Systems and then just have them Key it. They never have to touch the nonlinear animation setups in Houdini. You can then bring their animation clips into CHOPS and rework it into an automated Animation solution or some other kind of node network, etc, etc, etc.
I think if you are trying to use Houdini in the manner you spoke of then you are going about it all wrong. Houdini is the most powerful animation tool out there and I am not willing to give up funtionality and ease of use in light of simplicity. Aka Lightwave. Which is the reason I hate Lightwave because I can't do much in it without writing a mountain of code.
I do agree that it is always good to have access to a program through scripting and coding. I would like to see faster H-Script Instancing on that note too. Its a really cool feature.
Cheers,
Nate Nesler
Well honestly I hate Maya's rigging so to me that would be really bad thing to copy Maya. I think Maya's character system was out dated and was a good reason for them buying Motion Builder. XSI has a better Rigging and Character Animation System than Maya. I love Houdini's Rigging system over Maya. The reason why is because I can make a complex rig in Houdini without having to do alot of scripting and having to write alot of Maya SDK. Also they way I can build node networks and turn them into OTL's gives me a ton of flexibility with their rigging. Maya breaks down at the node network when you go to duplicate it because rarely maya duplicates the actual nodes and you get something different so that means you have to then write a custom MEL script to go through the steps to recreate that node network. I also ran into something fun in Maya with rigging and that was how many relationships it could handle at once on a character rig with the skin binded in Maya 4.5 they had the nasty problem of if you added the 5th relationship the entire character and all of its parts exploded across the screen and there was no undo because undo was frozen. So you just had to load back up your last save. Also in 4.5 they had the problem of twisting the skeleton through itself everytime you loaded the file. Which of course was no way like you saved it and would then throw off all of your expressions with your custom made graphic icons. So then you would have to go back and redo all of your expressions everytime you saved. In houdini I can create as many relationship as I want and Houdini has never blown apart on me. The only thing I have ever have to worry about is running out of ram or nodes corrupting. Locking nodes seems to keep this from happening but it does tend to happen from time to time. I am not really sure why.
I would also disagree with you on the modeling. You don't model in the tradional since in Houdini like the other packages you listed except for XSI. XSI also bogs down if you do not freeze its node chains. aka delete history in maya or lock nodes in houdini then select all inputs and delete the nodes. In houdini I model completely differently with a totally different workflow then I do in Silo, Modo, XSI, ZBrush, Blender, Lightwave, Maya, etc, etc, etc. Houdini can be your fastest modeling tool once you setup your own system of OTLs. You can use all aspects of the application to model with. I do have to say the fact that you can't go to quads but you do triangles and the fact their is not a quad clean up tool is a problem and does kind of throw the money wrench into the gears so to speak. Also the no displacement shader baking on Houdini Polys is a problem too.
I don't really think the Animation side is a problem because you don't have your animators that don't know anything about 3D actually get under the hood. You have TDs, Riggors, Smart Technical and Artistic persons to it with OTLs and Node Systems and then just have them Key it. They never have to touch the nonlinear animation setups in Houdini. You can then bring their animation clips into CHOPS and rework it into an automated Animation solution or some other kind of node network, etc, etc, etc.
I think if you are trying to use Houdini in the manner you spoke of then you are going about it all wrong. Houdini is the most powerful animation tool out there and I am not willing to give up funtionality and ease of use in light of simplicity. Aka Lightwave. Which is the reason I hate Lightwave because I can't do much in it without writing a mountain of code.
I do agree that it is always good to have access to a program through scripting and coding. I would like to see faster H-Script Instancing on that note too. Its a really cool feature.
Cheers,
Nate Nesler
- bugzpodder
- Member
- 12 posts
- Joined: July 2005
- Offline
- digitallysane
- Member
- 1192 posts
- Joined: July 2005
- Offline
I generally agree with MatrixNAN on the workflow “features”. However, for the things I would like to use Houdini for (broadcast, short deadline work) all those very efficient features are becoming irrelevant at the end of the job: the rendering is slow. The built-in shaders are incomplete, and, as I read on odforce, some of them are buggy. In this perspective, XSI is far ahead. Practically, for a broadcast job, all the time you might gain because of the advantages of Houdini will be lost when it comes to rendering. It is practically impossible to use area lights *and* motion blur in a project and expect reasonable render times (this is something usual with XSI). XSI renders faster with moblur than Houdini without. For me this is the only *real* showstopper when it comes to Houdini.
I know there are many professionals on this forum working on feature films and using mantra with various custom shaders who say mantra is very fast and efficient. For me, out-of-the-box, Houdini Apprentice is much slower in the lighting/rendering area than XSI.
Dragos
I know there are many professionals on this forum working on feature films and using mantra with various custom shaders who say mantra is very fast and efficient. For me, out-of-the-box, Houdini Apprentice is much slower in the lighting/rendering area than XSI.
Dragos
- peliosis
- Member
- 175 posts
- Joined: July 2005
- Offline
I don't think it's just as easy as saying one thing renders faster than the other.
MR is a raytracer so no doubt it raytraces faster than mantra.
But use some textures (eliptical filtering enabled), heavy displacement, bump mapping, put a few hundred of lights, everything in a scene with thousands of instances, increase AA to “production” values and then test.
Everything depends on what you are rendering, but from my experience of 3 years in xsi, MR is not the best option when it comes to fast, flicker free rendering.
I rarely used bump maps and displacement in xsi because of poor performance, I avoided night scenes because of the number of lights, and haven't used FG/GI in animation because it always was overkill. Things will probably change with 3.4.
Mantra has slow raytracing, ambient occlusion compared to dirtmap shader in xsi is unusable I don't have a great knowledge of it but definitely it can't be called just “slower”
As for motion blur, most of the time it's better to use reel smart, especially in broadcast when speed is more important than absolute subpixel perfection.
So if I had an architectural animation with 2 weeks deadline, no renderfarm I'd definitely avoid MR.
cheers
Peter
MR is a raytracer so no doubt it raytraces faster than mantra.
But use some textures (eliptical filtering enabled), heavy displacement, bump mapping, put a few hundred of lights, everything in a scene with thousands of instances, increase AA to “production” values and then test.
Everything depends on what you are rendering, but from my experience of 3 years in xsi, MR is not the best option when it comes to fast, flicker free rendering.
I rarely used bump maps and displacement in xsi because of poor performance, I avoided night scenes because of the number of lights, and haven't used FG/GI in animation because it always was overkill. Things will probably change with 3.4.
Mantra has slow raytracing, ambient occlusion compared to dirtmap shader in xsi is unusable I don't have a great knowledge of it but definitely it can't be called just “slower”
As for motion blur, most of the time it's better to use reel smart, especially in broadcast when speed is more important than absolute subpixel perfection.
So if I had an architectural animation with 2 weeks deadline, no renderfarm I'd definitely avoid MR.
cheers
Peter
- terrybuchanan
- Member
- 31 posts
- Joined: Oct. 2006
- Offline
Thanks for the post.
Just finished reading the article (and got registered on VFX World for the first time at the same time also). A lot of speculation about the future in the article. Have any of the predictions come to fruition since it was written over a year ago now?
How about in the industries other than film, such as commercials?
Is Houdini even used much in the commercials industry? Is that the correct terminology, “commercials”?
Is it easier for an entry level 3D person to get the foot in the door in the large studios working on commercials rather than film, and then transfer once the skills-set is honed?
What is the difference between “Special Effects” and “Visual Effects”?
A newbie interesting in learning,
T
Just finished reading the article (and got registered on VFX World for the first time at the same time also). A lot of speculation about the future in the article. Have any of the predictions come to fruition since it was written over a year ago now?
How about in the industries other than film, such as commercials?
Is Houdini even used much in the commercials industry? Is that the correct terminology, “commercials”?
Is it easier for an entry level 3D person to get the foot in the door in the large studios working on commercials rather than film, and then transfer once the skills-set is honed?
What is the difference between “Special Effects” and “Visual Effects”?
A newbie interesting in learning,
T
-
- Quick Links