houdini 13 as general purpose dcc tool.

[super gnu]

i know this has been discussed before, and the general consensus has been houdini is not ideal for the kind of jobs youd generally use 3dsmax or maya for, and is really suited to TD's doing vfx.

ive heard its clunky for modelling, difficult to pick up, a bit crap for general animation, but extremely powerful, especially for particles, dynamics etc. however you need to have a real programmers brain to get a decent result from it.

having watched a few of the intro tutorials and demos for r13, and read that they are in the process of improving the ui to be more -standard- and that the modelling tools in particular have improved a lot recently, im wondering if these observations are still so valid for the latest releases?

youd think not looking at side effects website. definitely marketed as a full dcc package.

they have an excellent pricing structure, very enthusiastic development, lots of free training material, actual real customer service, and most important of all.. they are NOT AUTODESK.

ive let my subscription slide, and im at a crunch point since they decided to discontinue upgrades. max (whatever AD might say now) seems to be being gradually sent out to pasture, and if i want to continue to pursue top of the line 3d (in whatever industry) changes are gonna be needed medium term. so thats either maya or something else. maya has always given me a headache, it doesnt yet do much that you cant (at a push) do in max, and again im at the mercy of the least customer-centric company on the planet.


im considering that houdini might be a good one to learn, since it seems to have all the things they are now -not- adding to max, as long as i can find a good way to inter-operate between the two, or, eventually replace max altogether. thoughts?

[samuel_bubat]

I also became a big fan of Houdini. What I like about it, it´s so consistent. Everything works together especially regarding dynamics.
It´s not that you have 10 different tools like in max. It offers everything regrading dynamics you ever wanted, fluid systems, particles,
fracture systems, hair and fur, cloth, rigid bodies, softbodies and they all can affect each other.

Regarding modelling. It does some things better and some things worse than max. Procedural modelling is awesome for things like stairs,
rails or similar. For low level poly modelling it offers a lot less tools and those are also not that straight forward to handle because of
the node based system. If you inset, extrude, edit polys you´ll get a huge tree of nodes. I like max much better in this case.

The biggest drawback of course.. no Vray for Houdini. Mantra seems to be ok. I can´t really judge after the short amount of time I´m using Houdini
while using Vray for more than 10 years now. But it offers a lot less options expecially regarding GI.

About interoperability. It has an option to export to fbx as well as alembic. But I can´t test them due to limitations of the PLE.
But with ply2vrmesh you have an awesome option to convert houdini geometry directly to a vrayproxy. It needs a bit work within
houdini to get everything prepared but it´s worth it. I started a tutorial some weeks ago how to export geometry from Houdini
to Vrayproxy including animations, fluid meshes, multisub materials, attributes like velocity, speed cusp and other. Also note
that an animated Vrayproxy is capabale of rendering motion blur. And the performance even with huge Proxies so far is pretty good.
I ´ll let you know when it´s finished.

So after all. I don´t think in our case that Houdini will replace max in the long term. 3dsmax to me is still the best tool regarding
poly modelling and merging everything together for final rendering. But when I should get a job that requires dynamics or is suitable
for intensive procedural modelling Houdini would be my first choice. Also because you can just rent it for the time you need it.

[super gnu]

your work with getting houdini stuff into vrayproxies sounds very promising. yes, from the very first tutorial i watched (building columns and shattering them) i thought.. blimey those node trees would get enormous on a complex model... about 15 nodes just to model a simple column!

apparently there are options to collapse or hide the nodes.. but it did make me wonder.. also i wonder about the concept (wrt. modelling) that "you can go back and change anything, its totally procedural" but of course unless i misunderstood something, if you go back and make a change that changes vertex/face count etc, youll still break the whole node tree, exactly like in max's modifier stack. so not quite so magically amazing...


im liking the idea of keeping my current max version to handle modelling and final (lovely vray) output, and have houdini sitting in between these two steps to do all the sparkly stuff that is now being left out of poor, tired max's development (as everyone knows its just a tool for arch. vis..doh. )

of course that process could be simplified if they release houdini engine for max (not holding my breath) or by having vray for houdini, but im not gonna be pushing for that yet !.. i only downloaded the PLE of houdini yesterday

[samuel_bubat]

This is test I made. Rendered in max with Vray.

http://sorceress.netfrag.org/optix/s..._vrayproxy.mp4

[super gnu]

impressive. and this workflow works with the PLE of houdini? could be just the thing for my current personal project. assuming i can also bring animated geometry from max into the PLE too.

[samuel_bubat]

Yes, this was done with the PLE. You should be able to import animated geometry with fbx. This works in the PLE but I haven´t tried it.

[super gnu]

well please let me know when you have the tutorial finished.. a rather cheeky fluid-sim solution!! maybe by that time ill have progressed beyond opening houdini and spinning the persp. view around.

[Morbid Angel]

The way we work with houdini is its possible to export alembic from it, which vray can read directly through vray proxy. This puts the transfer of meshes at ease for us. Volumetrics are also possible through vdb, but that's still work in progress.

[super gnu]

man the more i read about houdini the more impressed i am.. deep as the ocean. but the shelves with preset nodes seem a really easy way to get some fairyl complex results fast without being one of those people born with nodes between the ears, or a phd in applied maths. (im too oldschool. i still find node based workflows more intimidating than i should.)

[Paul Oblomov]

Well, alembic has its own drawbacks. Like moblur f.e.

[Morbid Angel]

Well, alembic has its own drawbacks. Like moblur f.e.

Whats wrong with mb in that?

yes alembic is far from perfect. FAR!

[Paul Oblomov]

Only one frame of moblur there.

[samuel_bubat]

Well, vrmesh also seems only to blur one frame.

Btw. I realized that you actually can output alembic files with the PLE (from a file node).
So after seeing Vlados gazillion particle test I gave the particle export a try.
I´m really impressed from Houdini as well as Vray. The attached frame
was a simple particle system with turbulance from a wind node. The spawn rate is 10mil.
Not sure how Houdini calculates this but if this means 10mil. per frame this would be 500mil. particles.
The calculation and export of the alembic files was roughly 10min. for all 50 frames.
the last frame file is 1,5gb.
I used a Domelight with an HDR. Fixed sampling 4, Embree Raycaster.
Vray renders this frame in ~2min on a single machine (quadcore) Ram usage was ~3gb
So Krakatoa might have a new competitor

[super gnu]

hehe that is also.. awesome. seems they didnt lock down the PLE of houdini too tight. good call.

ive been hoping vray could start stepping on Krakatoa's toes.. one less super specialised plugin to buy.

but did i understand right.. 1.5gb a FRAME..!? thats rather weighty.

[samuel_bubat]

Yes, but I guess that´s pretty normal for a cache file with 500mil particles.
They are stored as particles and not meshes, also no attributes like velocity were exported.
So I guess unless you compress you can´t go much lower. I would think if you cache
files from Pflow they would be the same size. Hmm I think that´s going to be my next test...
10mio spawnrate with pflow

I also noticed that when you convert to proxy wich is only a single huge file, the viewport performance
is much better compared to abc files.