Like most Softimage users, I'm checking out Houdini and other options. Houdini obviously looks very powerful, but honestly I have not yet seen any examples of really good models built in Houdini.

I've checked out some modeling tutorial videos and all of them seem to start out with some basic modeling concepts & then quickly move to all the neat procedural stuff you can do with the model.

But honestly, I don't need to make a basic chair that can interactively have longer or shorter legs. I need to model a very detailed chair.

I understand that's not the Houdini way of doing things, but the fear I have right now is, everyone using Houdini seems to be in love with the idea of setting up very complex procedural models that can be adjusted later, but when you look at the final example as just a model, it's not very good. It seems like people will spend hours and hours building something that could be cranked out in 30 minutes, except in the Houdini version you can tweak the bevels and thickness of things later on.

Are there examples of very detailed, impressive models that were built entirely in Houdini?

I don't mean to sound negative about Houdini - it really does look like a fantastic tool. But I don't really see examples of highly detailed models built in Houdini.

-Paul

I think it´s fair to say that it is perfectly possible to build detailed models in Houdini…but not very frequent.

When proceduralism isn´t needed for a model it´s not uncommon to use another app for modelling and bring that into Houdini.

This might change sooner or later, but for the time being, this tends to be the situation.

Paul Griswold
It seems like people will spend hours and hours building something that could be cranked out in 30 minutes, except in the Houdini version you can tweak the bevels and thickness of things later on.

Indeed - that happens, at least to me, all the time

Good opportunity to run a modelling competition

With prizes for best model. The rules include; does not care about the proceduralism. Just the final model.

Trust me, I'm very impressed with Houdini. My other fear is, I get very enamoured with things like being able to go back and change things later. It's really cool for sure.

I could honestly see myself spending way too much time trying to build something just so I could show everyone how many parameters you could change later on, when a basic model would have sufficed.

The inner geek in me takes over, but ultimately I'm in business to make money, pay my bills & take care of my family.

-Paul

Keep banging on the tools, make otl's when you spot yourself doing the same thing over and over. Houdini makes spotting patterns in your workflow easy to do so take advantage of that! Don't get hung up on proceduralism (unless it's an obvious solution) and instead focus on just making the model and take advantage of the non destructive nature of Houdini's workflow.

I am by no means cranking out beautiful models with Houdini yet, but I have only been seriously studying the modeling side of it for a month. And in that time I have seen tremendous growth in my knowledge and capability with Houdini as a modeler.

A Houdini modeling contest is a fantastic idea!

Lyr
A Houdini modeling contest is a fantastic idea!

The competition could coincide with some of the revised modelling/viewport features:

1. It would bring attention to these new developments

2. Put those tools to the test in the real world.

I'm about 4 days in learning Houdini and one thing became perfectly clear to me last night.

Don't plan to go about using Houdini like Softimage. It misses the point entirely. I suspect there are some really good reasons existing Houdini experts don't miss the tools us exiled Softimage users have grown so accustomed to.

I'm still all for the devs putting some of that stuff into Houdini. But if they never did, I highly suspect I am not going to miss it before long.

Two examples that lead me to the above notions. 1. I just took a cube (without knowing much of anything about Houdini) and split it with the loop tool a few times, threw on some extra operations, then went into the network and deleted some of the earlier loop splits and surely enough, the end result was immediately broken and un-usable. I am sure their are nodes I could put into the pipe to fix such issues, but the point of it is, that just going about things in a traditional, Softimage way, where you are directly editing the model by hand, and at the same time ignoring the network making a straight vertical stack of nodes, isn't going to lead to an immediately procedural result and misses the point. 2. I have been diving into Houdini tutorials, and I'm seeing all new ways to make things and I think I am going to be a whole lot happier throwing out my old methods (and just loading up an old copy of Softimage when I really need them).

I'll leave it at that, I wish I could have said the above with less words.
This series is by far the best resource I've found for a Houdini beginner IMO: http://www.sidefx.com/index.php?option=com_content&task=blogcategory&id=202&Itemid=361 [sidefx.com]

Generaly nodes that work on explicit point/edge/.. Ids won't be ‘procedural’

eg procedural edge loop split would need another object/location&vector defining plane of cut then node calculating edge IDs that cross that plane and feeding result array of IDs into edge loop split node.
And then all following nodes would need to be procedural like this too.

'Hardcore' modeling node stack in H is exactly like in SI
Sometimes you can go back and fix, especialy when fixing bevel height etc parameters that won't add/delete prims.
But most of time you freeze modelling after a while (when soft slows down) and go on.

Hey Piotrek! Good to see you here too!

I'd be interested to hear your take on Houdini.

-Paul

I think Derek really hit it right… it is a very different approach for modeling, but it can be very successful (and in many cases more successful that the usual “interactive” modeling workflows)

Unfortunately I cannot post examples from my work, but we have used houdini to model photorealistic assets in minutes VS weeks using procedural methods (i.e. hundreds of heads, broadcast graphics, terrains, etc).

Often our methods involve a mix of the procedural and the interactive, but we are very careful at choosing where one makes more sense than another. For example, seeing people model fishscales in mudbox by using stencils is good, but I feel I could make a much more sophisticated fishscale method in Houdini. Likewise, I feel like i can model a “fishman” faster in zbrush than in houdini

one thing that actually surprises me is that Houdini is often compared to tools like maya and max for its modeling, but I would consider tools like maya and max (and even soft to some extent) not the first place where I would go start a model. I tend to prefer hardcore modeling programs such as zbrush and 3D coat for interactive modeling. hard surface modeling is another story, but sometimes those things are great candidates for procedural methods

This one is an oldie and not really crazy detailed (this was before zbrush/mudbox came on the scene heh) all done in Houdini; pretty much box modeling with polysplitting/polyextrudes/edgeflip. Some things that help speed up the workflow is keybinds for your most common tools and RMB on the parms/toggles and setting new defaults (next time you drop down the SOP it will have the toggle set so you don't have to keep setting it). Depending on what you're modeling you don't HAVE to make it procedural. Some things can be purely procedural and some lend themselves to good old point pulling.

And if you're having any issues in viewport please submit bugs with your specs (Help->About Houdini->Show Details) so they can be fixed.

In the past, houdini has been mostly associated with high-end movie stuff, but I could see houdini being awesome for doing stylized low-poly modeling for video games. It could be especially good for independent designers/studios where there are only, say 1-5 people working on a whole game and not the resources to make a game filled with extremely detailed models.

A bit like pixel art, you don't need thousands of pixels to express a concept artfully. You just need fine control over each pixel.


12++