I'm looking quite closely at viewport interaction in Houdini and how we can make it awesome. Any and all suggestions are welcome.
Thanks for the warm welcome… I've been lurking (and doing some work) in Houdini for several months so by now I've got a few notes for sure. Granted I'm really new to Houdini so there's a spectacularly high probability that there's something I'm missing… and if that's the case then I'd be quite grateful to be pointed in the right direction.
This first note isn't so much related to viewport interaction, but in the beginning I found that the workflow for editing groups of components (created by selection) was not terribly obvious. Like it was unclear (just by looking at the UI) how to go back and add or remove selected components from a group after the fact (other than by typing in numbers). Eventually I found the “~” key and figured out the “enter-key-to-commit” thing, but I certainly missed the simplicity of the old SI “add/remove to cluster” - I'm still not entirely sure what the tilde key really does or if there's a gui equivalent to it, although I did see what command that hotkey maps to…
And while I know that a group can be generated by any operator, and that it's maybe not as simple as putting in a or icon in the active toolbar, I'm pretty sure there's gotta be a good design solution for that somehow (I still don't remember how I found the tilde key thing, probably on the very-helpful forums or odforce or something).
In general, I think the component selection tools themselves could also benefit from some love… the tools I've missed are “select parallel edge loop”, “select border components” (it's there, but maybe broken for prims in 13.0.314?), “select border edges” (like open mesh edges, the kind you'd want to cap in order to close off a mesh), and while the existing edge loop tool “works” it's still a little funky sometimes.
Even though I'm doing mostly procedural tasks in Houdini, I still often need to do things like shape a mesh surface or model a quick and dirty collision or control object and it would be nice if modeling were just a little easier than it is.
In SI proportional modeling was a big, huge timesaver. and while softPeak and softTransform are really nice, it would be even slicker if I could then drop down a weightmap and paint the effect back down without leaving the 3d viewport (is there an existing way to do this without doing it manually?).
Another minor annoyance is (transient) windows getting lost underneath the main viewport window. So if I'm browsing for a SOP path in a parameter field or doing that “edit parameter interface” thing and I make an errant click somewhere, then I have to go looking for the window I just opened which is now lost beneath the main UI. Anyway it would be nice if windows with “accept” or “cancel” just stayed on top until I was done (like the help window).
The other big one is being able to visualize custom attributes more easily. I've taken to building a set of default display attributes (scalar/vector.. can we have an axis for matrices?) so that I can drop down a “bind export” (if I'm say in a VOPSOP) to debug an attribute at a specific point in the node graph, but I wonder if there isn't maybe a better way.
Oh and passthrough (or “dot” like in Nuke) nodes would really help with node graph legibility and accidents when “drop on a wire” is active (which I like, but is really dangerous for big crazy graphs). Really miss those passthroughs…
Thanks for listening… It's been interesting and fun spending time in this new package.