I almost pulled an all-nighter, so if I sound a bit bitter, you'll know why (I'll try not to)
The reason is, I tried to understand Houdini's capturing workflow, to a level deeper than “superficial”, not TD deep as I am not one, but deep enough to serve my posing/short-anim needs, as a creature/character artist wannabe.

Thus far I have nothing to say about the three capturing methods available at the start, other than it's cool to have many options - choose w/e suits your case.

But then it comes the down in trenches work of editing the weights. I often heard that the starting weights are very close to the final state. This is only partially true and for the cases that it is true, it's working great from the get-go for regions like elbows, wrists, knees etc and not too bad for more proximal and medial ares also. But then, if you reach the head/mouth or have a creature/humanoid that's slightly different in a few important ways (like having big adductors very close to each other, or big hanging lips/ears, I can go for much longer with examples) you can forget about fiddling with capture regions. In fact, you can forget them and accept that you're not going to escape painting weights for hours or even days, the moment you have a realistic character (be it a “mere” human) that makes use of corrective shapes (another painful subject) coupled with blended algorithms (dual-q and linear) and is expected to look and deform as close to reality as possible.
Judging the current state of capture editing by how well it performs on a cartoon character from the get go is a mistake.

Another big show stopper is the painting performance. I don't think I need to go into detail here.

So back to the painting workflow, but stopping first at the capturemirror node - I have read the docs description of the three options of the Behavior algorithm and I still don't think I fully understand them, as they're worded in a very abstract manner, but what do I know, maybe there's no other way. Anyhow, without proper testing on some simple examples, which I didn't have time for so far, there's no way I'll fully grasp their workings. What I did however notice while working on this creature I've spent the night with (in Houdini) is the default option Leave Unmirrored Captures gave me a lot of trouble by producing spikes (unnormalized weights). Copy Unmirrored Captures gave me consistent and predictable results with the exception of one region where a ridge between the halves was present and couldn't get rid of it. It turned out to be not the fault of the Behavior option, but some bone assignments that had no business in that area.
Michael told me that he uses the first option “Leave…” and works for 99% of the time. That maybe so, but if you're reading this and have the same issues I had, consider switching to “Copy…”

Painting (Capture Layer Paint) and numerically adjusting the weights of individual or groups of points (Capture Override)
I fail to see the reason for the 2nd node existence. Now, I'm talking theory here, not taking shots at SideFX or anyone, I did this before (“I fail to see the reason…”) and I might've been misunderstood - these tools are always created in a technical env. constrained by limitations and a mass of tools/techniques you can build on top of and you work with what you have in the present. Time has passed and now modern approaches have been devised so it's time IMO to address some of these areas, capture editing being one.
Capture Override for example, being accessed only when you're at Scene lvl, from the shelf, is not something anyone thinks is great, I'm sure. But beside this awkwardness, there's the question of why do I need a separate node for manually entering numerical values for points? Maybe this is the XSIer in me talking and I'm failing to consider something specific to Houdini, but the painting node paints with numerical values after all, so why not have a “Spreadsheet” button on its PPG? Have a rebel point that just won't be picked by your brush? Simply select it, fire up the Spreadsheet and enter the desired value. This is the workflow I'm used to and any unnecessary complication (like having a separate node that works only under some inconvenient setting) is a huge downgrade and an overhead bump in performing simple editing tasks.

Sorry for the wall of text. I know some of these have been discussed and IIRC, someone from SESI (don't remember who) said it's an area about which they're not happy with. I just thought I'd leave my input here and maybe come with some insights that have not been put on the table before.
Cheers!

I'll be frugal and just use this thread for a more down-to-earth issue - what's the current workflow for adding bones to a capture solved mesh? I just realized that I'd be better off if I added some bones in a certain area, but wish to keep the overall capture weights.
So far the search engine directed me to various threads old and new (mostly old) that didn't ultimately help me. Is there a simple, or at least sure-to-be-working, method of adding bones to an already “skinned” geometry?

pretty sure there is a ‘regular’ way…but I can't remember it

what I usually do is make my bones, then use a captureLayerPaint and add the new region into the Capture Region field, then paint to add influence…

I'll hunt around and try to find something ‘official’

I tried re-applying the capture using the shelf tool hoping that “append” method and the fact that I've added the new bones in bundle I use to feed to the “individual bones” method, would do it. I guess I'll try your way of painting influence for each bone.

Hello and nice to see you again Mc Nistor. Sorry to semi hijack your thread. I need help with my first ever Houdini skining.
The source is an FBX object that Imported, rescaled separated into different parts. It's a cardboard box for which I need to animate its unfolding.
So I created some bones and used the “capture geometry” shelf tool, and that's it, I immediately got the polygonal object disappearing and these errors at the geo level.
The first error says “points have no assigned weights” but isn't this what the shelf tool is supposed to do?

Picture attached.

Thanks
Gz

Hey man,
W/O sharing the actual scene file, it's hard for me to know what's happening. If you can't share it here, let's go private and if you can't share it at all, maybe someone else will figure it out just from that image.

I'm going to take a guess and say that your cardboard box has no ‘thickness’ and so the solid embed isn't able to build much of a tet mesh
I'd switch to ‘proximity’ method and try again.

Thanks a lot to McNistor through Skype and to Goldfarb. The problem indeed was the biharmonic option, once I chose Proximity, all went well.
Then I started to edit the weight using the shelf tool “Edit Capture Weights”. Some problems I found:
- The doc says one should first select the points to correct, then click the “Edit Capture Weights” shelf tool, but that's not correct, as selecting this tool removes the selection and leads you to reselect the points.
- Once the spreadsheet is displayed, I started to reselect the point but … bugger!! Grow selection, using SHIFT+G, isn't available in this mode.
- In the spreadsheet column titles cases : some bright colors in the background prevents me from reading the bone name. One workaround is to RMB on these title cases, then “edit bone color”, then choose a darker color. But I beleive Hou should allow to keep bones in a bright color if the user needs it, and automatically darken these chosen colors for the background of the title cases.

All three of these points are known and have RFEs logged to address.
thanks.