Maya to H11 rendering

[br1]

Hi !

I'm doing some research for a featurette that I'll probably have to render in the next months.

Everything is done in Maya until we reach the point of rendering. I want to avoid using mental ray as much as possible as it gave us a lot of trouble in our last production. I see two choices here. We do not have the budget for renderman or 3Delight licenses, but could afford using V-ray as we already have V-ray licenses and would just need a few workstations with the Maya plugin.

My other option would be to use Houdini and Mantra for rendering but I have some questions on the subject.

As it will only be used for rendering, I guess Houdini Master is not required. I was also thinking that one license per lighter would be enough and we will not need to buy anything else as we can render on as many nodes as we desire. Is that right ?

About transferring the data from Maya to Houdini, are there tools that will help us ? I had issues with fbx crashing Houdini. I also tried the bgeo writer from houdinistuff.com but was having problems exporting a rigged character with multiple meshes. It could be scene related, so if anyone has experience or better workflows I'm intrested.

It will be a stereoscopic film. Is it easy to transfer stereoscopic cameras from Maya to Houdini ? And is Mantra optimized for stereoscopic rendering like Renderman is (speedup when rendering the other eye)?

some things will still be rendered in Maya (fluids, particles, etc) as we do not have the budget for Houdini Master licences. I want to be sure that the renders will match between both packages (aspect ratio, subdivisions, etc). Any advice on the subject ?

That's it for now, maybe more will come to my mind later. Thanks for reading !

Bruno

[haggi]

Mantra has no stereoscopic features like renderman, you will have to render your image twice. Indeed, you can get as much mantra licenses as you need with a commercial version of houdini.

If you are not very familiar with mantra rendering and shader building in Houdini than it will take some time to switch, more than you will think.

In maya its a bit simpler to connect your nodes to a network, e.g. plug a textureFile node into the diffuse color of a lambert shader or simply use only the red channel for bump. In houdini it is better but often a bit more work.

Just for interest, what are the problems with mentalray that you have?

[abvfx]

there are a few decent stereoscopic features but i dont think there is any optimization going on or a way right out of the box to import a two camera setup.

This page on the stereo rigs might be of some help.

http://www.sidefx.com/docs/houdini11.0/nodes/obj/stereocamrig

[LaidlawFX]

Mantra can do stereoscopic rendering just like renderman, we have it set up at work, dome style even. For the normal user it's not press and go though it takes a little settup. You should be able to get in touch with sesi to get the run though.

You should hire some one to help setup a pipeline for you. I've done a bunch of maya to houdini work, and at the simple level it can be done pretty easy, but for a production it can get complex. It's well worth the money. I am in the process of setting one up now for our company and depending on the amount of back and forth it can get hard quick. It depends on how flexible you want your connection between packages. Also I don't mean you need a R&H/DD/Pixar level pipeline, you can make a very simple one that is run on a few script in python/mel from maya to export, and a couple otls in houdini to import, but take the few weeks to smooth out the workflow. It will make crunch time a lot smoother.

Escape is a good option, use .obj and a good naming convention for geo, fbx to get cameras over(you'll need a custom script for stereo), develop a shader in houdini that is the same as a maya shader (you can do a one to one options setup for all options in maya shader). Lighting is better in houdini so no need to transfer that.

hope that helps

[br1]

Mantra has no stereoscopic features like renderman, you will have to render your image twice. Indeed, you can get as much mantra licenses as you need with a commercial version of houdini.

If you are not very familiar with mantra rendering and shader building in Houdini than it will take some time to switch, more than you will think.

In maya its a bit simpler to connect your nodes to a network, e.g. plug a textureFile node into the diffuse color of a lambert shader or simply use only the red channel for bump. In houdini it is better but often a bit more work.

I'm already familiar with shading and lighting with Mantra, although yet I didn't go too deep in shaders, often having the desired result with the already existing shaders and a bit of tweaking. I have the feeling that the mantra surface shader will do for most of the asset we have to render. It'a mainly characters with a bit of sss.

Just for interest, what are the problems with mentalray that you have?

Well, the list is rather long but just to name a few problems. Note that we where using mental ray for Maya, not a standalone version.

Memory issues when rendering on the farm.

Frames that will render in the viewport will crash when rendered on the farm.

Having to manually adjust subdivisions on each shot.

No optimisation for stereo rendering forcing us to recalculate final gathering for each eye.

Subdivided geometry crashing mental ray.

Render problems if skinned characters are to far away from zero.(subdividing them brings some noise to the subdivided mesh).

Problem with render passes forcing us to split our scenes in different versions to be able to output all our passes.

That's what I have in mind. Maybe things have changed since Maya 2010 but I'm not too optimistic.

[haggi]

Mantra can do stereoscopic rendering just like renderman, we have it set up at work, dome style even. For the normal user it's not press and go though it takes a little settup. You should be able to get in touch with sesi to get the run though.

Then I was completly wrong, sorry. I was told by Jeff Wagner that there was a lot of discussion about stereoscopic optimisations in mantra. And they decided to do it the good old way without any reusing sample tricks how it is done in PRman. Seems I didnt pay attention to what he said.

[haggi]

That's what I have in mind. Maybe things have changed since Maya 2010 but I'm not too optimistic.

A lot of things have changed since maya2008. e.g. rendering with bsp2 and instancing as been improved a lot. Now you can rendern hundreds of thousends of instances spread with particle instancer without wasting too much memory. Passes, final gather has improved as well as other elements. We rendererd a whole animated movie with mentalray and most with rasterizer and it worked well.

[br1]

A lot of things have changed since maya2008. e.g. rendering with bsp2 and instancing as been improved a lot. Now you can rendern hundreds of thousends of instances spread with particle instancer without wasting too much memory. Passes, final gather has improved as well as other elements. We rendererd a whole animated movie with mentalray and most with rasterizer and it worked well.

Well all the problems I talked about where with Maya2010, not 2008. I don't know if things did change a lot between Maya2010 and Maya2012(I was busy with other things ^^). I still think that for characters with displacement and motion blur work mental ray is not the best option.

Edited June 17, 2011 by br1

[br1]

Mantra can do stereoscopic rendering just like renderman, we have it set up at work, dome style even. For the normal user it's not press and go though it takes a little settup. You should be able to get in touch with sesi to get the run though.

You should hire some one to help setup a pipeline for you. I've done a bunch of maya to houdini work, and at the simple level it can be done pretty easy, but for a production it can get complex. It's well worth the money. I am in the process of setting one up now for our company and depending on the amount of back and forth it can get hard quick. It depends on how flexible you want your connection between packages. Also I don't mean you need a R&H/DD/Pixar level pipeline, you can make a very simple one that is run on a few script in python/mel from maya to export, and a couple otls in houdini to import, but take the few weeks to smooth out the workflow. It will make crunch time a lot smoother.

What kind of person should be hired ? Do you think it can be handled with just python and mel ? I also have the feeling that it well worth it's money and if things run smoothly for our featurette, the pipeline will probably be re-used or enhanced for the full-feature film coming after.

Escape is a good option, use .obj and a good naming convention for geo, fbx to get cameras over(you'll need a custom script for stereo), develop a shader in houdini that is the same as a maya shader (you can do a one to one options setup for all options in maya shader). Lighting is better in houdini so no need to transfer that.

hope that helps

I don't really understand what you mean when you say "Escape is a good option". Can you clarify ?

And we will not shade or light anything in Maya, so we don't want to translate Maya materials, we'd rather do them all in Houdini and reassign them automatically based on naming conventions.

Thanks for your inputs !

[LaidlawFX]

Then I was completly wrong, sorry. I was told by Jeff Wagner that there was a lot of discussion about stereoscopic optimisations in mantra. And they decided to do it the good old way without any reusing sample tricks how it is done in PRman. Seems I didnt pay attention to what he said.

It's possible Antoine was helping work with the dev of it. I understand the concepts of it enough to know what it is and how it functions, not do it myself. So I wouldn't say completely wrong, directly off the shelf we may have to wait for h12.

[LaidlawFX]

What kind of person should be hired ? Do you think it can be handled with just python and mel ? I also have the feeling that it well worth it's money and if things run smoothly for our featurette, the pipeline will probably be re-used or enhanced for the full-feature film coming after.

I don't really understand what you mean when you say "Escape is a good option". Can you clarify ?

And we will not shade or light anything in Maya, so we don't want to translate Maya materials, we'd rather do them all in Houdini and reassign them automatically based on naming conventions.

Thanks for your inputs !

Make sure you have a houdini/maya generalist that knows programing well. If they have already made a converter or two for cameras and .obj exporters for maya, and some otls for importing into houdini that would be a good asking question. There are a few scripts boucing around, the net you can use, but the person hired should be able to understand every line of code in those small scripts. I know it was part of one class at AAU, so a SCAD person should have no problem with it, too.

Second this person should have a focuses on lighting, shading, and rendering, which would be harder for the houdini side as oppose to the maya side. It's an easy cross over, from houdini to maya lighting most kids oout of school light with may. This is a good time during the summer to find some one that may have worked on Hop, or a couple of the other productions that have rendered with Houdini in the last year for a feature with H11, since, these back end artist aren't needed now for the feature cycle. Commercial work is cool, they may have used H11 more possibly, but commercials kind of drop the people off at more random times of the year.

Thirdly a person who sells themselves as Houdini Pipeline person would be the best case, but you will probably be paying a lot of money for them. I would say it's better to train up, unless you hire one of these guys for a short/medium stint, and have them train some one up. I am sort of in this position now, which is why I say it.

Houdini "Escape" is a good option cost wise if you are just going to have a bunch of lighters, shader artist, and render folk. You do no need the features of houdini master, I've been doing all of my work in escape with regards to this side of the pipeline for the last two months, master prior. The only thing I open master for lately, has been to grab some icons to make my otls look prettier. Suffice to say I haven't done any "non-shader/sop effects" lately.

As for naming convention you can set up your maya scene exactly like your houdini scene in sops. UVs, group names, normals, and hierarchy(in group context) transfer over. As for characters you'll just have to output obj sequences, though it may be recommended that you have a display version that is low poly. Suffice to say have a proxy display, that is separate from your render. Houdini's OGL with poly counts is not as forgiving as mayas, with real time play back especially with viewport 2.0.

[br1]

Thanks for all your precious information Ben.

Just out of curiosity, has anyone used alembic to transfer the data between both packages ?

http://code.google.com/p/alembic/

We'll maybe have a few days to test things, but feedback from experienced users is still welcome.

Bruno