SI and Houdini

   26832   38   1
User Avatar
Member
20 posts
Joined: March 2014
Offline
Thank you for this warm welcome. I have decided that going back, exclusively, to Maya as my primary package is not something that my brain is willing to tolerate (I left it as a primary tool for good reason). I started learning Houdini yesterday (via the learning edition and a few good first tutorials) -I am impressed and I feel a well deserved spirit of rejuvenation and curiosity growing again, like being back in a good school once more.

Even though it is officially dead, Softimage is still going to be a staple in my bag of tricks for some time to come and I'm sure I will continue to use Maya from time to time as I never fully stopped using it, but as far as new dedicated effort on my part goes, I'm going to be putting that entirely into mastering Houdini. Thank you for, well, being here, it is greatly appreciated and it means a lot to me. Best of wishes to all.
User Avatar
Member
5 posts
Joined: March 2014
Offline
jordibares
If you look at the Softimage to Houdini transition guide I am already diving into workflow elements bit by bit and soon will discuss how to manage assets and pure production workflow so stay tuned.

Oh yes youre really helping us a lot. I already saw all the pdfs from the first time i was here.
Thank you so much.
User Avatar
Member
5 posts
Joined: March 2014
Offline
I dont know if there is another way to do it fast in Houdini i just need to search more. If there tell me please.

Well the point is.

1. Looking at videos of mantra rendering point clouds, first of all, now i really know how fast was (is) ICE particle sims even with million of particles you are able to press play a see results, mantra way is replicating cached sims from what i saw on vimeo videos.
2. Besides that maybe it can be very cool to have something similar to emrpc or exocortex fury, very very fast particle renderers. self shadowing, gi, motion blur, dof, apart from replication if you want, and almost instantly.

Was playing with a file from Anthony Martin(search Vimeo) and this was taking like 30 seconds per frame SIM + Render non cached 2880*1620.
Shadow quality set to low.
https://www.youtube.com/watch?v=t-ogoCtWNLA [youtube.com]

And this one like 12 seconds per frame SIM + Render non cached 1920*1080
Shadow quality is kind of below normal.
https://www.youtube.com/watch?v=QodftoU3y-M [youtube.com]

Both of them render with Fury2 in Dell XPS Desktop I7, 16 RAM and a normal 2gb ATI card.

Maybe i have to clarify that those renders were only for learning purposes.

That would be very very nice.
User Avatar
Member
11611 posts
Joined: July 2005
Offline
Hi there,

I threw together a simple sort-of comparison for you; some advected particles in a fluid sim, and then pointReplicated to bulk it up for free.

The longest part of the render is sending the geometry to Mantra - on my poor home machine it still rendered pretty fast – just a few seconds.

Does this help at all?

Thanks!
Jason

Attachments:
lotsOfPoints.hip (872.7 KB)
lotsOfPoints.jpg (226.6 KB)

Jason Iversen, Technology Supervisor @ Unity Technologies x Weta Digital
also, http://www.odforce.net [www.odforce.net]
User Avatar
Member
204 posts
Joined: Nov. 2013
Offline
hey jason,

thanks for the hip file. unfortunately there is no shader applied to the points.
do you mind to elaborate what kind of shader you used?
thanks a bunch

sebastian
EDIT:

never mind. the camera was in the same position as the light.
thanks again!
http://www.sekowfx.com [www.sekowfx.com]
User Avatar
Member
11611 posts
Joined: July 2005
Offline
Sorry, I probably saved the file in a hurry. As you've probably realised anyway, there is a default shader applied in the case of missing shader assignments. Anyhow I hope the file provides the simplest demonstration of point rendering, shadow mapped lighting (if you're after sheer speed, this is the quickest way) and that style of simulation.
Jason Iversen, Technology Supervisor @ Unity Technologies x Weta Digital
also, http://www.odforce.net [www.odforce.net]
User Avatar
Member
4189 posts
Joined: June 2012
Offline
Alan Torres
2. Besides that maybe it can be very cool to have something similar to emrpc or exocortex fury, very very fast particle renderers. self shadowing, gi, motion blur, dof, apart from replication if you want, and almost instantly.


You might be also interested in this particle rendering comparisons from Symek:

“Particle Rendering Deathmatch!! Arnold vs. Vray benchmark test.”
http://forums.cgsociety.org/showpost.php?p=7707775&postcount=20 [forums.cgsociety.org]

Thread:
http://forums.cgsociety.org/showthread.php?t=1134058 [forums.cgsociety.org]
User Avatar
Member
304 posts
Joined: May 2006
Offline
Did Symek ever share his scene? I would like to benchmark with it…
Javier Meroño
FX TD.
User Avatar
Member
4189 posts
Joined: June 2012
Offline
That would be cool, this was the thread:

http://forums.odforce.net/topic/18957-mantra-particles-rendering-test/ [forums.odforce.net]
User Avatar
Member
1390 posts
Joined: July 2005
Offline
Netvudu
Did Symek ever share his scene? I would like to benchmark with it…

Nope, because I messed it up during tests (wasn't saving changes) but will try to find it later this week.
User Avatar
Member
304 posts
Joined: May 2006
Offline
great!….and also abusing of your patience, when you check it, could you leave it set up with the ink-like rendering (volumes)? it´s a setup I always wanted to see myself.
Javier Meroño
FX TD.
User Avatar
Member
1755 posts
Joined: March 2014
Offline
I was thinking about opening a thread “Does Houdini have?”, but then I thought this one is good enough.
So I'll start wit hthis:

does Houdini have a way of referring to another objects coordinate sys. for viewport transform purposes?
For example in Softimage if I want to move an object based on anothe'rs local coord. sys, I go ‘Reference> pick object reference’ in a menu which allows me to transform an object with the gizmo aligned to the picked object local coord. sys.. \
If it's not clear let me know.

Attachments:
ref.jpg (91.3 KB)

User Avatar
Member
187 posts
Joined: Oct. 2015
Offline
+1
As a fresh new indie user I try to see if Houdini is sufficiently user friendly for “direct” modeling.

McNistor, I also really interressed if there is a way of referring to another objects coordinate sys. for viewport transform purposes.
User Avatar
Member
1755 posts
Joined: March 2014
Offline
Hey julca, welcome to the dark side.

Houdini seems to have an “Orientation Picking” but it doesn't seem to be very useful as far as I tested it. It's either buggy or if it's working as intended, it surely isn't helpful.
If you are to pick another object as reference (which you really can't, since it asks only for points or edges), the selected object jumps to the “reference” position for some reason - if I wanted that I'd use snap, “match translation” or simply transform it manually.

On top of these things, Houdini only aligns the gizmo to certain things, but it doesn't change the coord. system. For example you can't edit the global coord. of points directly (only using an attribute wrangle).
Translations are really a big problem right now and unfortunately the veteran Houdini user-base is not interested, or at least they don't seem to be as it's probably outside the scope of the work they do in Houdini. Just guessing, of course.

And there are other problems. Hopefully they will be solved sooner rather than later.
User Avatar
Member
187 posts
Joined: Oct. 2015
Offline
Hello !

Thanks you for your explanation, I also really hope they will solved this in a near future.

I think that really small improvements around modeling interaction can radicaly change modeling feeling and productivity (I give me some days before discuss about more in depth).
User Avatar
Member
1755 posts
Joined: March 2014
Offline
julca
I think that really small improvements around modeling interaction can radicaly change modeling feeling and productivity

I couldn't agree more. I said this many times in the past - improving the tools (bevel, extrude, etc) is very important, but if interaction (selecting, transforming, etc) is not top-notch, then those poly operations become just another set of nodes to be used in the netview in a procedural manner. The procedural philosophy must of course be taken into account as that's the heart of Houdini, but should not be the only thing driving the development and improvement of new and respectively old workflows.
User Avatar
Member
187 posts
Joined: Oct. 2015
Offline
And that's exactly that's I want to say !
When I say small improvements around modeling interaction it is about the development effort that can make all modeling much more user friendly.


In other words, I really agree with you that interaction (selecting, transforming, etc) must be “perfect” to be really productive (that's huge repeating tasks in a day work !).

In the same way, I don't think that's so hard for the dev to reach this level, example :
- add judicious transient hotkeys for snap/“transform handle” (aka “pivot” in xsi)
- make transform tool more flexible : match transform - pick polygon/edge reference, etc.
User Avatar
Member
1755 posts
Joined: March 2014
Offline
julca
match transform

Starting with H15 there's a feature for matching any of the transforms (see small button beside each value box). What's missing is a “match all” button or function (like hold shift and click any of those buttons).
User Avatar
Member
187 posts
Joined: Oct. 2015
Offline
McNistor
julca
match transform

Starting with H15 there's a feature for matching any of the transforms (see small button beside each value box). What's missing is a “match all” button or function (like hold shift and click any of those buttons).

Yes ! That's good to know !
  • Quick Links