I know people are stupid writing that Houdini is a technical, hyper complicated application and not using it and not being interested at all.
I found houdini to be an easy solution to create and control very advanced effects and objects.
I have one question though…is there anybody making arch viz in houdini?
What are mantra capabilities on rendering architectural objects with their specific problems?
How does mantra speed compare to others? Is it a good idea to render architecture with mantra and are ther any examples in the web?
If anybody have some more experience in this field I would be grateful for some help.
I read something about vops complexity adn lack in ease of use.
(I use xsi rendertree and find it very convenient)
Thanks
Peter
Architecture rendering in Houdini.
16144 12 3- peliosis
- Member
- 175 posts
- Joined: July 2005
- Offline
- old_school
- Staff
- 2540 posts
- Joined: July 2005
- Offline
Unfortunately there are no users that can be considered Architects involved in building design and visualization using Houdini that I know of.
That aside, I have seen Houdini used to build large virtual 3D citiy sets for both commercials and feature film. Hopefully others can comment on this with example images.
In a previous lifetime I was involved in the architectural community and am always tempted to do 3D vis in Houdini. I don't see any reasons why it wouldn't be suitable. As a matter of fact, you could probably build a collection of really smart HDA's to build the bits of buildings procedurally.
Rendering: Speed-wise, Mantra is right up there in speed when doing straight raycasting and raytracking. Motion blur is very fast (thanks to reys). As for the GI rendering options, the parameters relating to GI have to be tweaked somewhat before you get adequate speed in rendering.
Yes it is a good idea to use Mantra for rendering architectural scenes.
As for VOPs, if you are familiar with rendertree in XSI, then VOPs is just a bit of a side-step. Should be very straight-forward transition.
The main differences between RenderTree and VOPs is that VOPs have to be pre-compiled before you can use the shader, known as a SHOP. This is as simple as RMB on the vop network and choose make shop. The link is live. Changing the vop network is instantly reflected in the referenced shop. In XSI, the rendertree is the interface to the shader that is used for MentalRay and the shader is compiled on-the-fly.
That aside, I have seen Houdini used to build large virtual 3D citiy sets for both commercials and feature film. Hopefully others can comment on this with example images.
In a previous lifetime I was involved in the architectural community and am always tempted to do 3D vis in Houdini. I don't see any reasons why it wouldn't be suitable. As a matter of fact, you could probably build a collection of really smart HDA's to build the bits of buildings procedurally.
Rendering: Speed-wise, Mantra is right up there in speed when doing straight raycasting and raytracking. Motion blur is very fast (thanks to reys). As for the GI rendering options, the parameters relating to GI have to be tweaked somewhat before you get adequate speed in rendering.
Yes it is a good idea to use Mantra for rendering architectural scenes.
As for VOPs, if you are familiar with rendertree in XSI, then VOPs is just a bit of a side-step. Should be very straight-forward transition.
The main differences between RenderTree and VOPs is that VOPs have to be pre-compiled before you can use the shader, known as a SHOP. This is as simple as RMB on the vop network and choose make shop. The link is live. Changing the vop network is instantly reflected in the referenced shop. In XSI, the rendertree is the interface to the shader that is used for MentalRay and the shader is compiled on-the-fly.
There's at least one school like the old school!
- jason_iversen
- Member
- 12654 posts
- Joined: July 2005
- Offline
I'd like to open a modeling workflow discussion off this, if you don't mind
I think while Houdini and Mantra are fully capable of managing and rendering architecture, there is a sore spot when it comes down to modeling. Oftentimes geometry imported from architectural software, which can often be very badly constructed geometry gleaned from atuotraces of floorplans and such, the resultant geometry you are left with is often very bad. Take a city model, for instance, how do you want to handle it? All the geometry in one object? Or hundreds/thousands of seperate objects (each a building or connected mesh)?
There may be dozens of objects which have reversed faces. There is no good way to select 50 objects and Reverse SOP them all. Or divide, or any modeling operation. Keeping them in one object would be incredibly slow - reversing faces would have to build the entire geometry again and when you're talking about 600k polys - ouch. Even hitting “u” and “I” to pop into SOPs to model on something causes a huge slowdown since the network pane cant really handle large numbers of nodes efficiently and the context change and take 3-4 seconds. I really believe SESI would benefit from changing this paradigm a little and allowing modeling operations across multiple selected objects. It needs to be done in todays age of bring Houdini into the general (not-only-effects) market. Right now it's not really viable to model on large scale scenes in Houdini. Its perfect for building up scenes from scratch, but attempt to fix and refine a large dataset and you'll be better off doing this step elsewhere for now.
On the rendering topic, I believe mantra has what it takes to render beautiful images and you'll find its handling of memory to be very efficient and fast.
I think while Houdini and Mantra are fully capable of managing and rendering architecture, there is a sore spot when it comes down to modeling. Oftentimes geometry imported from architectural software, which can often be very badly constructed geometry gleaned from atuotraces of floorplans and such, the resultant geometry you are left with is often very bad. Take a city model, for instance, how do you want to handle it? All the geometry in one object? Or hundreds/thousands of seperate objects (each a building or connected mesh)?
There may be dozens of objects which have reversed faces. There is no good way to select 50 objects and Reverse SOP them all. Or divide, or any modeling operation. Keeping them in one object would be incredibly slow - reversing faces would have to build the entire geometry again and when you're talking about 600k polys - ouch. Even hitting “u” and “I” to pop into SOPs to model on something causes a huge slowdown since the network pane cant really handle large numbers of nodes efficiently and the context change and take 3-4 seconds. I really believe SESI would benefit from changing this paradigm a little and allowing modeling operations across multiple selected objects. It needs to be done in todays age of bring Houdini into the general (not-only-effects) market. Right now it's not really viable to model on large scale scenes in Houdini. Its perfect for building up scenes from scratch, but attempt to fix and refine a large dataset and you'll be better off doing this step elsewhere for now.
On the rendering topic, I believe mantra has what it takes to render beautiful images and you'll find its handling of memory to be very efficient and fast.
Jason Iversen, Technology Supervisor & FX Pipeline/R+D Lead @ Weta FX
also, http://www.odforce.net [www.odforce.net]
also, http://www.odforce.net [www.odforce.net]
- old_school
- Staff
- 2540 posts
- Joined: July 2005
- Offline
- hoknamahn
- Member
- 398 posts
- Joined: July 2005
- Offline
I really believe SESI would benefit from changing this paradigm a little and allowing modeling operations across multiple selected objects. It needs to be done in todays age of bring Houdini into the general (not-only-effects) market.Really good idea.
f = conserve . diffuse . advect . add
fx td @ the mill
fx td @ the mill
- peliosis
- Member
- 175 posts
- Joined: July 2005
- Offline
Thanks for your replies.
Since I don't use models made by any architecture studio for they are just as good as noticed above, I was not concerned about modelling.
I count on houdini modelling and I see a model of the building made of not many objects (10-20?) There are large amounts of identical objects in buildings so I didn't consider it a problem.
When modeling a building I begin with cleaning dwg's no matter in which software I model. Then…with the wireframe of positioned curves(most often IGES file) I have a ghost to fill up with geometry.
XSI also becomes very slow with large amounts of objects, it handles a million of polys in one piece, but if there are more than a hundred of 10000 poly objects…oh my.
I'm very at the very beginning with houdini, and I think It can be very fast in modelling architecture. It would be nice to have a living form, and when a customer says they had to change the half of the project you just change some curves and values and everything seamlessly updates…free from slicing cutting moving points rebuilding from scratch.
I'm also obsessed with the thought of an animated self building construction, everything working like an organism.
So am I on a dead end way? would houdini be overwhelmed by architecture projects and painstakingly cease to cooperate?
If yes please turn me around
I was mostly concerned about rendering.
I grew on 3ds max 1-4 and it was a robust renderer no matter how complicated geometry it had to render.
Now when I work with mental ray I find it hard to complete an animation before deadline and I wondered how does mantra compare to other software when it comes to raytracing (glossy reflection, GI, area shadows…)
We may switch to modelling, but please tell me more about mantra if you know.
thanks
Peter
Since I don't use models made by any architecture studio for they are just as good as noticed above, I was not concerned about modelling.
I count on houdini modelling and I see a model of the building made of not many objects (10-20?) There are large amounts of identical objects in buildings so I didn't consider it a problem.
When modeling a building I begin with cleaning dwg's no matter in which software I model. Then…with the wireframe of positioned curves(most often IGES file) I have a ghost to fill up with geometry.
XSI also becomes very slow with large amounts of objects, it handles a million of polys in one piece, but if there are more than a hundred of 10000 poly objects…oh my.
I'm very at the very beginning with houdini, and I think It can be very fast in modelling architecture. It would be nice to have a living form, and when a customer says they had to change the half of the project you just change some curves and values and everything seamlessly updates…free from slicing cutting moving points rebuilding from scratch.
I'm also obsessed with the thought of an animated self building construction, everything working like an organism.
So am I on a dead end way? would houdini be overwhelmed by architecture projects and painstakingly cease to cooperate?
If yes please turn me around
I was mostly concerned about rendering.
I grew on 3ds max 1-4 and it was a robust renderer no matter how complicated geometry it had to render.
Now when I work with mental ray I find it hard to complete an animation before deadline and I wondered how does mantra compare to other software when it comes to raytracing (glossy reflection, GI, area shadows…)
We may switch to modelling, but please tell me more about mantra if you know.
thanks
Peter
- anon_user_40689665
- Member
- 648 posts
- Joined: July 2005
- Offline
- old_school
- Staff
- 2540 posts
- Joined: July 2005
- Offline
I too think about how you could automate the construction of buildings. Taking different bits and mixing them together to see what you get.
As for the rendering of large datasets, all of the renderers based on the Reyes (renderman) model render using buckets including Mantra. The renderer only needs to pull in as much data as it needs to complete the current bucket it is rendering.
This is why both Mantra and Renderman can load in insane amounts of data. It only has to load in that portion of the scene. If you run in to memory trouble, reduce the bucket size.
Mantra does raytracing just fine. Pretty fast. As for the GI stuff, you could search the od-force forums to see about rendering GI effects. There are some neat things you can do with point clouds to create GI effects.
As for the rendering of large datasets, all of the renderers based on the Reyes (renderman) model render using buckets including Mantra. The renderer only needs to pull in as much data as it needs to complete the current bucket it is rendering.
This is why both Mantra and Renderman can load in insane amounts of data. It only has to load in that portion of the scene. If you run in to memory trouble, reduce the bucket size.
Mantra does raytracing just fine. Pretty fast. As for the GI stuff, you could search the od-force forums to see about rendering GI effects. There are some neat things you can do with point clouds to create GI effects.
There's at least one school like the old school!
- MatrixNAN
- Member
- 405 posts
- Joined: July 2005
- Offline
Hey Peter,
You know you can also use Mental Ray if you aren't comfortable with Mantra although if you know Air then you will have a really easy time doing Mantra. They are going to be almost identical. Mantra is based around Renderman just like Air is. So … you shouldn't have any trouble. Well like one of the other posts I put up in dealing with architecture. I would create and Otl that was say for a window. Then load pieces of the frame that could be loaded in as one SOP and which could then be scaled, rotated, translated, with shops already applied, and you could choose a style of windows that relates to certain objects. That would allow you to minamize your rendertimes and generate buildings much faster. Because those pieces that are loaded into the window OTL can all be instaced and configured off of loaded in pieces. Check out the candle OTL tutorial if you have not already but it will give you a real good idea of what you can do with OTLs. Then just think of how you can apply that to Architectural Visualiazation. Houdini does instances pretty fast in Mantra.
Cheers,
Nate Nesler
You know you can also use Mental Ray if you aren't comfortable with Mantra although if you know Air then you will have a really easy time doing Mantra. They are going to be almost identical. Mantra is based around Renderman just like Air is. So … you shouldn't have any trouble. Well like one of the other posts I put up in dealing with architecture. I would create and Otl that was say for a window. Then load pieces of the frame that could be loaded in as one SOP and which could then be scaled, rotated, translated, with shops already applied, and you could choose a style of windows that relates to certain objects. That would allow you to minamize your rendertimes and generate buildings much faster. Because those pieces that are loaded into the window OTL can all be instaced and configured off of loaded in pieces. Check out the candle OTL tutorial if you have not already but it will give you a real good idea of what you can do with OTLs. Then just think of how you can apply that to Architectural Visualiazation. Houdini does instances pretty fast in Mantra.
Cheers,
Nate Nesler
- madjestic
- Member
- 325 posts
- Joined: July 2005
- Offline
- goldfarb
- Staff
- 3463 posts
- Joined: July 2005
- Offline
http://sidefx.vislab.usyd.edu.au/houdini_video/by_topic/general/index.html [sidefx.vislab.usyd.edu.au]
- madjestic
- Member
- 325 posts
- Joined: July 2005
- Offline
- mrw558
- Member
- 1 posts
- Joined: June 2021
- Offline
-
- Quick Links