Katana Infos ?

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

THREAD CLOSED
 
Thread Tools Search this Thread Display Modes
  06 June 2013
Dorian >
Thanks dude this sound like a nice tricks ! it reminds me the scene builder i was using in one of my previous employer ( you know the one i'm talking about ). A nodal Python UI which is able to build scene base on text file is i guess one of the best options for a maya lighting workflow. if you don't have some C++ soldier army

SYmek >
Always a pleasure to read your very informative post man !
While i'm at 10000 km of your Houdini Expertise, i definitly feel from the small i know that H can definitly be a fantastic weapon as a base of a lighting workflow. Unlike in maya you don't have to code your external scene builder , Houdini IS the scene builder and offer you impressive set of features by default ... thus my question of H vs Katana !

From what have been said on Katana Price / Feature wise, it looks to be more oriented for the 150-2000 employees companies ...

But as the world that interest me is more the one of the 1-150 employees Maya or H are the 2 remaining options i guess ... So i was trying to know what the best tool has to offer , in order to see if such a workflow can be done cheaper with some H or Maya customisation.

I'm quite impatient to discover the Arnold Integration inside H , it will maybe offer a VOP/VEX level of control ? and a Mantra / Arnold hybrid workflow looks exotic as a full year in Bahamas ...

Cheers

E
__________________
emfx.fr
 
  06 June 2013
Originally Posted by earlyworm: That's fine if all your working with is Houdini, but the thing is these days you've got to be able to play nice with other applications, departments and companies.

The thing (I'm not sure I'd call it hype) about Alembic is that it's an open format. Before that every large vfx studio around had (or still do have) their own unique file format for storing 3d data.


Oh, I totally agree! Besides openness Alembic is superior to Bgeo* (mostly the old one) for other reasons too (extensibility, disks space and speed, although - pipeline wise - not everyone is happy with an idea of single huge cache file).

I was just saying that deferred load is something Houdini workflow is aware of because keeping everything outside a scene file is very natural for this app.

Also, let my state clearly, I don't argument against Katana, this is simply an opinion from the point of view, which doesn't have many issues Katana was designed to solve .


* - Biggest problem was code dependency though. SESI has published example library but it never was fully functional or maintained (though used for old time Maya plugin). Likewise linking your code with HDK wasn't an option too. In fact that's pretty much the same story with other stuff, like great 3d texture format existing for +10 years, their own 3d LUT, deep images, channel clips and a few others goodies largely wasted in Houdini's land due to license restrictions. Paramount here is PIC format with anything you need from an image container mostly unused for decade.
 
  06 June 2013
Originally Posted by SebKaine: It reminds me the scene builder i was using in one of my previous employer ( you know the one i'm talking about ). A nodal Python UI which is able to build scene base on text file is i guess one of the best options for a maya lighting workflow. if you don't have some C++ soldier army

Owww yeah...

Originally Posted by SebKaine: From what have been said on Katana Price / Feature wise, it looks to be more oriented for the 150-2000 employees companies ...

It's just something I notice more and more:

Less you have access to the price, more time/dev/td you need to make it work.

I suppose there is some invert examples to this... Maybe Guerilla?

Originally Posted by SebKaine: But as the world that interest me is more the one of the 1-150 employees Maya or H are the 2 remaining options i guess ... So i was trying to know what the best tool has to offer , in order to see if such a workflow can be done cheaper with some H or Maya customisation.

I completely beggining to H (and love it! ) but as a Rendering TD: More I use it, more I realize it could be a "katana for the poors". I mean. Houdini has a very good support of alembic reading (I realize few weeks ago Maya alembic reader was completely missing important options, like "output by name" for example...) so you could use it as an "assembling application" to get every information from the "exterior" (geo/materials/assignations, etc...) reconstruct all procedurally and render.

The problem is if you are not familiar with Mantra (I'm not yet). If you have Vray licences you could see Vray Standalone Data access (understand this by: "Vray standalone execute a python script telling him to reconstruct and render the scene from scratch"). But as you are on a <150 peoples, I think learn a little more Mantra should be very efficient (H Mantra mentors are hard to find).

Originally Posted by SebKaine: I'm quite impatient to discover the Arnold Integration inside H , it will maybe offer a VOP/VEX level of control ?

Hahaha. I don't know about Houdini2Arnold team. What I can guess is that Mantra is so well implemented to H that it will be hard for Arnold to find it place, except if it became a "standard" (which is not the case yet but they are working on that! ).

Maya actually never had a good renderer implementation (IMHO) by I don't think it's a renderer problem: The need to use nodes for everything is a completely crazy approach in lighting (just look at the graph you have when you have light un/linking, render layers, etc...) This can quickly bring to thousand of connection for actually do something very simple. The only thing that work well with nodes is material dev.

Originally Posted by SYmek: Oh, I totally agree! Besides openness Alembic is superior to Bgeo* (mostly the old one) for other reasons too (extensibility, disks space and speed, although - pipeline wise - not everyone is happy with an idea of single huge cache file).

+1 but I don't think it's an Alembic problem.

I mean, if you don't want one single file for all caches, you can separate them as you which. Char Prim, Char Second, Set FG, Set BG, etc... One of the strongest point of Alembic if the speed to open/read. You could take an advantage of this to have few (not so much) separated caches. If you have well named and coherent structure in every abc file this is not a problem to use later.
__________________
My blog (fr)
 
  06 June 2013
Quote: I completely beggining to H (and love it! ) but as a Rendering TD: More I use it, more I realize it could be a "katana for the poors". I mean. Houdini has a very good support of alembic reading (I realize few weeks ago Maya alembic reader was completely missing important options, like "output by name" for example...) so you could use it as an "assembling application" to get every information from the "exterior" (geo/materials/assignations, etc...) reconstruct all procedurally and render.

The problem is if you are not familiar with Mantra (I'm not yet). If you have Vray licences you could see Vray Standalone Data access (understand this by: "Vray standalone execute a python script telling him to reconstruct and render the scene from scratch"). But as you are on a <150 peoples, I think learn a little more Mantra should be very efficient (H Mantra mentors are hard to find).


Well we use Mantra for lighting a lot of FX elements and it leaves vray behind in the ability to render massive data sets.

Quote: I'm quite impatient to discover the Arnold Integration inside H , it will maybe offer a VOP/VEX level of control


What do you mean like the ability to write shaders. ?

B
 
  06 June 2013
Originally Posted by Narann: Hahaha. I don't know about Houdini2Arnold team. What I can guess is that Mantra is so well implemented to H that it will be hard for Arnold to find it place, except if it became a "standard" (which is not the case yet but they are working on that! ).


Well for FX artist Mantra is more than enough, the problem is that where i use to works there is a huge gap beetween the approach of a good FX guy and a good Lighting guy ! Lighter are mainly pragmatic artist that just want to go fast without bothering too much with shaders / technics / etc ... V-Ray and Arnold are quite popular cause they offer a good trade beetween ease of use and quality. I'm afraid that lighter where i work will not agree to dive in Mantra ... Arnold is not perfect but at the end the philosophy of quick setup and long render is pretty efficient , so if i have to light cars / liquids / parfum / big outdoor env / mechanical elements ... i think Arnold would be a clever choice ...

Originally Posted by mr Bob: What do you mean like the ability to write shaders. ?


Well something like the Mantra Shader, you have all your basics components , but you have VOP access inside the shader if you want to modify it or connect any attribute you like. What would be really cool is to be able to get a similar workflow than in Mtor with the Delux shader, where you add the shading component you need. While the delux doesn't conserve energy at the time i use it , it would be cool if this additive operation keep the PBR concept. (exactly like Maxwell Do ... I Love Maxwell !)
But 2 other area that would be great is to have :
- a very rich library of basic prebuild shader like in Maxwell
- the ability to communicate shader / light setup beetween Maya and H in a Fluent way like in maxwell with the external shader editor

EDIT : when i re-read my post i'm starting to ask to myself if Maxwell is not the way to go for lighting ...
__________________
emfx.fr

Last edited by SebKaine : 06 June 2013 at 07:17 AM.
 
  06 June 2013
@mr Bob
I've not been very clear but SebKain has answer.
Yes, if you want to have a good integration in Houdini, you should think about data access, this is one of the true force of Mantra integration (IMHO). You have access to everything directly in the shader.

Originally Posted by SebKaine: EDIT : when i re-read my post i'm starting to ask to myself if Maxwell is not the way to go for lighting ...

Because if you say Arnold is slow, you will realize Maxwell is far far far slower.
__________________
My blog (fr)
 
  06 June 2013
Originally Posted by Narann: Because if you say Arnold is slow, you will realize Maxwell is far far far slower.


Well Dorian , this is what everybody use to say about maxwell , it is very very very slow ! I'm not very good at rendering But we start to use maxwell where i work and we have one maxwell master that really know what is he talking about ... and actually it can be pretty fast for production , with displace , motion blur and DOF less expensive than in V-Ray on certain shot ! Again i'm out of area of knowledge here . But be careful with evreybody saying maxwell is not usable in production. For Commercial work it can be a nice choice. Especially when Maxwell 3 will arrive ...

check this all is computed the guy just use post computed tweaks like you would do in lightroom with a RAW :
https://www.youtube.com/watch?v=087-FdCyUwg

For one frame lighting design purposes its IMPRESSIVE !
__________________
emfx.fr
 
  06 June 2013
Originally Posted by SebKaine: V-Ray and Arnold are quite popular cause they offer a good trade beetween ease of use and quality.


I'm not sure if this is correct as an implicit comparison to what Mantra offers these days. Once Arnold will be available for Houdini, we will see a migration (or not). Otherwise it's hard to say people prefer Arnold over Mantra for simplicity. Leaving all quirks of hybrid engine aside, you can use Mantra exactly as Arnold with brute force path tracing. We do such tests from time to time trying to recreate in SI what was rendered in Houdini. It actually hasn't happened in so far that Arnold was faster or looked any different/better AFAIR (although these test are not specially rigorous).

Last edited by SYmek : 06 June 2013 at 11:53 AM.
 
  06 June 2013
Originally Posted by SYmek: I'm not sure if this is correct as an implicit comparison to what Mantra offers these days. Once Arnold will be available for Houdini, we will see a migration (or not). Otherwise it's hard to say people prefer Arnold over Mantra for simplicity. Leaving all quirks of hybrid engine aside, you can use Mantra exactly as Arnold with brute force path tracing. We do such tests from time to time trying to recreate in SI what was rendered in Houdini. It actually hasn't happened in so far that Arnold was faster or looked any different/better AFAIR (although these test are not specially rigorous).


The trade off with Arnold is being 2hrs later you get a nice render !
 
  06 June 2013
you should take a look at atomkraft for nuke (https://atomkraft.hk/nuke), it's a renderer (3deligth)integrated in nuke. perfect for look dev and comp.

an you also have Creation: Stage beta (http://fabricengine.com/creation-modules/stage/)
currently support arnie and vray and soon renderman.

http://vimeo.com/61635020
http://fabricengine.com/creation-mo...ting-rendering/
__________________
popol va au cirque ....
 
  06 June 2013
@ mr Bob + SYmek >
After comparing both Arnold and Mantra you looks to find Mantra as fast (faster ?) and as good as Arnold.
Mantra token are absolutely free so you can have a farm with 100.000 CPU for 0$. So why big boys buy Arnold instead of Mantra. I would say that on top of that Mantra has a shading language while Arnold don't (in his generic form ... ). It's hard to believe that skilled people will prefer to spend money on a meal while they could have the same one for free ...

@popol
Thx for the links !
__________________
emfx.fr
 
  06 June 2013
Originally Posted by SebKaine: @ mr Bob + SYmek >
After comparing both Arnold and Mantra you looks to find Mantra as fast (faster ?) and as good as Arnold.
Mantra token are absolutely free so you can have a farm with 100.000 CPU for 0$. So why big boys buy Arnold instead of Mantra. I would say that on top of that Mantra has a shading language while Arnold don't (in his generic form ... ). It's hard to believe that skilled people will prefer to spend money on a meal while they could have the same one for free ...

@popol
Thx for the links !


Faster ? I really don't get why would you would think of comparing them like that.
A tool is as only as good as the people who set it up I wont name shows but Ive seen Katana set up so badly it got dropped on the next show at company A, meanwhile Katana at company B works like a charm to a degree, but still is so slow and clunky on bigger shots it becomes frustrating for lighters.

Who makes the decisions ? , the CG sup running the show gets the input. Depending also on what the show requirements are we will use any render engine. Mantra gets currently used for Volumes and to render massive data sets like a cloud of instanced geometry. All fx is done in Houdini.

B
 
  06 June 2013
Originally Posted by mr Bob: Faster ? I really don't get why would you would think of comparing them like that.
A tool is as only as good as the people who set it up I wont name shows but Ive seen Katana set up so badly it got dropped on the next show at company A, meanwhile Katana at company B works like a charm to a degree, but still is so slow and clunky on bigger shots it becomes frustrating for lighters.

Who makes the decisions ? , the CG sup running the show gets the input. Depending also on what the show requirements are we will use any render engine. Mantra gets currently used for Volumes and to render massive data sets like a cloud of instanced geometry. All fx is done in Houdini.

B


@ Mr Bob >
You're absolutely right ! ... but only in the canvas of vfx for motion pictures in a big companie. In the world of commercial in small shops (where i live now) :
- you are the pipeline maker (cause there is no R&D )
- you are the TD problem solver if things go wrong
- you are your own supervisor ( most of the time )
- you have N/rand(3-5) to do a job that would ask N time in reality
- you have no renderfarm and sometimes you lunch all on your local machine ...

In this world of CG guerilla such things as:
- how many time take a frame to render on this renderer
- how many time take a shot to setup on this renderer
- how time do i need to integrate the engine in the workflow
are quite essential ...

But i agree with you guys Mantra is a perfect engine for throwing all FX. question is when you send your mesh to the lighters what are now the cool options, especially when lighters are force to work extremely fast. I have never work in a place that use Mantra as a main lighting engine. Arnold has this big advantage to be extremely easy to understand, while mantra is more on a Prman Base that has evolve to a hybrid Prman / Arnold ...

I love Mantra and i don't need as a FX guy to go on other engine, but i'm not sure that mantra is still the best choice to light Cars / Large env / Glass / Parfum ...
__________________
emfx.fr

Last edited by SebKaine : 06 June 2013 at 07:22 AM.
 
  06 June 2013
Originally Posted by SebKaine: Mantra token are absolutely free so you can have a farm with 100.000 CPU for 0$. So why big boys buy Arnold instead of Mantra. I would say that on top of that Mantra has a shading language while Arnold don't (in his generic form ... ). It's hard to believe that skilled people will prefer to spend money on a meal while they could have the same one for free ...


Mantra is not free. You need hscript and Houdini to work with it, which is expensive (from big studio perspective you need site license most probably). Secondly, it binds you to Houdini and SESI, which might not be the best idea either. This is all quite complicated. Mantra could be exactly the same as Arnold (or better), and still it might be wiser for, say, ILM to work with Solid Angle, at least until SESI won't declare it's going into renderers market (making separate division, selling Mantra without H etc). At the moment Arnold as written by 16 people (?), while SESI perhaps little bigger, is focused entirely on FX, and it's clearly intentional that SESI doesn't go into renderers - highly difficult market, like it has never wanted to compete with PRMan (for many complains about lacks in Mantra backend, SESI usually says: you can do it in Houdini, Mantra doesn't have to know how to do it).

Like it or nor, Mantra is positioned as a fx renderer AND as a main renderer for smaller facilities, which are important for SESI, as a source of competent people (slightly simplifying - "software culture" I should say) because lack of them historically was the main reason on Houdini's problems.

At the end there is nothing free. You either buy technology or produce one. If you buy it, try to pickup the best people, and diversify sources, because you hardly ever find the best in everything in a one place.

On top of that Arnold might be simple better from Mantra in brute force path tracing as a product of very focused development (it's definitely worse when it comes to flexibility), but I haven't seen evidence of it yet.
 
  06 June 2013
Thanks for your precise answer SYmek , it does make sense !
__________________
emfx.fr
 
Thread Closed share thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright 2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 02:13 PM.


Powered by vBulletin
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.