lewis_T I really don't follow your aversion to even having everything in SOPs, with dopnets inside it.
Right, just my preference and the first thing that popped to mind in answering the question. For me, the Vellum Solver, the pyro sparse solver, are a great step in that direction. And for sure, there are many things about Houdini I wish I had time to learn. Super fun app.
2. HDAs are overkill. I would much rather have Nuke's Toolsets so I can save groups of nodes without going into build-a-tool world, i.e., make it easy to save and reuse node chains, like when you put down a Component Builder node set. That nifty pscale trick Simon did in the Titan cables video? Make it trivial to to grab, reuse and share that set-up.
You could just grab the nodes to a shelf.
I've known you could do that for a million years and it never occurred to me to do it. Thanks!
lewis_T You can do all your work in one level inside SOPs.
Lewis, Nor sure I follow you. For example, if I build a crowd agent in SOPs, I then have to put down a DopNet, dive inside and start building the crowdsolver network. It's this diving in and jumping up I wish Houdini didn't have. I want to build it all on one flat level.
1. There is no work flow. You spend too much time jumping around. I can traverse up and down a Nuke tree for hours without having to stop and think about where a node or parameter is. I change a thing, I see the result and keep going. This also makes it easy to hand another artist a Nuke script without much explanation - they just read down the tree. Houdini scripts are usually spread out over too many levels and contexts to make them easy to share and understand. Solaris, Vellum and sparse pyro are going in the right direction but let's do it everywhere.
2. HDAs are overkill. I would much rather have Nuke's Toolsets so I can save groups of nodes without going into build-a-tool world, i.e., make it easy to save and reuse node chains, like when you put down a Component Builder node set. That nifty pscale trick Simon did in the Titan cables video? Make it trivial to to grab, reuse and share that set-up.
3. Axe the OP. I personally hate the OP vernacular. /mat and /stage tell me something rather than being random acroynms my brain has to stop and decipher. Again, the paper cuts. It feels related to that whole thing where people talk about Prisms once in awhile. It's a blast from the past but really, it isn't helpful, especially when you start explaining things to newbies. And LOPS / Solaris / USD / stage / is...I don't know, something.
4. The help docs. The help docs have long felt like they were written at a command line prompt. No judgment but for VEX you get this kind of thing: <type> attrib(<geometry>geometry, string attribclass, string name, int elemnum) which is cludgy at best. Rarely did you get a real world example, e.g., pos = point("defgeo.bgeo", "P", 3) which is all you really wanted. They're getting better and now I see screen grabs for node trees occasionally (e.g., crowds). I remember when they added them for ForEach loops and I was blown away. Too often though, its just "this sets up this attribute," without real help as to what it does in the context of that node. I know a bunch of old school users will bang away at how great the docs are but it feels people without a CS background often aren't well-served by the docs. CGWiki, PixelFondue, Entagma fill in the gaps but that they are so necessary to understanding Houdini is also telling.
5. Who's on First. I second the AttributeCreate example someone posted above. I would add the nontrivial time it takes to sort out how to get an attibute from one place, e.g. a VEX wrangle, to another, a parameter in a node can get confusing as all hell. Also what you can feed into a parameter (e.g., expression, VEX, pattern-matching, etc) can trip you up. It feels like moving attibutes around could be much cleaner but I'm sure there are very good reasons it is not. I know we're in experienced user territory now but looking back at the time it took to learn it, it feels like it didn't need to be that hard.
These are some of the things that have made Houdini a stuggle for me to use and advocate for it. It's a great DCC but this is the first year in 10 I've seriously considered not reupping my license. It's powerful but pretty unique and with so many really great and very accessible tools (Unreal, Axiom, Blender, Nomad, etc) it feels like the time to final frame is pretty high.
For background, I've used Houdini since v.11, mostly for effects on indie films. I started in 3D with Shake, then Nuke, then Houdini so I can't compare to Maya or 3DSMax. My core s/w is Houdini, Nuke, Substance and Nomad on a Mac (a whole ∆ painful topic).
Thanks for the chance to add the feedback.
TL/DR: Houdini's ratio of Productive Work to Why Isn't It Working? is very low because Houdini lacks a cohesive UX.
Is there a way to turn off the viewport Hints (the Python state hints) permanently? I hide them but they reappear when I leave the tool and return. Bit of a pain as they take up a decent amount of space.
Is your paramater pane open in the network view (by hitting P)? I've noticed that my viewport behavior can be erratic, including as you describe, when the parameter pane is open, and closing it sometimes solves the problem. I'm also on a Mac (desktop) but haven't tried with my Mac laptop yet. Also, I haven't filed a bug report b/c I'm able to reproduce it every time.
Continued development of a brand new renderer; dev and integration of a render delegate for third party renderers; Continued USD integration; Revising PDG/TOPs to be more accessible; Rewriting character animation tools; Deepening use of VDB/sparse tools; Continued dev of robust professional modeling tools; countless bug fixes and the usual VEXification and optimization of nodes…
would be a heavy lift for any company but
all while working from home, dealing with the continued Metal/Cuda/OoenCL and GPU/CPU whackiness during a period when its anyone’s guess when film production will crank up again…
I have no idea how you schedule against that and will be happy if 18.5 hits anytime before Hallowwen.
I'm thinking about going to a dedicated Windows or Linux machine for Houdini, but keeping Mac to run all the other tools. I'm going to wait until the fall, see what AMD does, etc.
The biggest downsides for Mac for me are:
1. SideFx incorporates third-party tools, e.g. Substance plug-ins, Alice Vision, etc that run on Windows but not Mac;
2. Apple's shader API, Metal, is a non-trivial hurdle for SideFx (and all renderers) to adopt;
3. Apple hates Nvidia for childish reasons which keeps RTX, other Nvidia cards out of the Apple ecosystem, including the otherwise-amazing “Pro” machine;
and
4. Apple is transitioning to ARM-based machines. This will create its own cyclone of issues.
as for RAM management, Apple does that really well, insert usual Apple caveats, mileage may vary, etc.
Nice renders! Let us know when your video hits YouTube, love to check it out.
Midphase I'm going to make a YouTube video soon about it …
Here's a still from a project that I've been working on lately…using MOPs to instance the buildings, then adding VDB's for fog/atmosphere and displacement to add more interest to the geometry. Render time in HD was around 3 minutes, but I'd probably push the quality settings higher for a final render:
Midphase Even if 18.5 or 19 offer nothing more than bug fixes
that's what daily builds are for and can be accelerated by submitting bugs and reproducible cases I'd definitely not like if SideFX copied some other companies and just slapped new version on the bunch of bugfixes
100% this. Foundry's Modo v10 was a paid upgrade for bug fixes. They were quite brutal about it. I would be for a version somewhere after Solaris / Karma / TOPs, etc all get settled in that was devoted to quality, speed, etc with less emphasis on new features. Much like Apple did with its Snow Leopard release (and should do again IMHO).
And I am using PCs for rendering anyways.
