applying orentations to bones

[meshsmooth]

I have a spine and would like to apply the blended rotations on several blend objects to the bones

So

[meshsmooth]

the file

apply_orentation_to_a_bone.zip

[meshsmooth]

Ok is it because nobody knows about chops there is no response?

I am working around my limeted knowledge of chops and I am about to make a new object that has most of the properties of a bone but is based on a blend object, so it can be constrained.

So what features make a Houdini bone what it is? This is my list, any surgestions

* must have a length and another bone conected to it will start at the end of the bone

* 3d bone like geometry along that length (so it look like a 3d bone)

* The ability to capture regions

* be able to be attached to its parent and receive rotations and position (this is where I deviate because what we are trying to achieve is bones with constraints that can be applied to them)

* receive IK... Now these bones will not be able to have IK solvers applied to them but this is where you use normal bones.

This is a theory at the moment. So I am just trying to flesh it out.

I believe by the time I finish my rig an am happy with it I should have a strong idea on Houdini

[danteA]

I emailed you my example.

[meshsmooth]

I didn

[meshsmooth]

I have just had a close look at the solutions and there cook times.

All cook times are with the whole test case the nulls that provide the orientations and the bones receiving the rotations make up .36 ms of the time

"Dante's chop solution", cook time: 1.74 ms .... Minus bones etc.: 1.38

This is the slowest abut potentially the neatest solution. This could be made into a chop subnet OTL and be very net indeed.

"Dante's blend object solution", cook time: .70 ms .... Minus bones etc.: .34

fastest but causes the most visual clutter in the network, most fiddly to create.

"Roberts BoneBlend OTL solution", cook time: 1.17 ms .... Minus bones etc.: 81

Made after reading an explanation of Dante

apply_orient_dante2.zip

[edward]

What kind of speed are you expecting? 0.34 ms looks good from here.

[meshsmooth]

Well there are 1,000 milliseconds in a second, that makes it only 40 milliseconds if you chop it up into 25 frames.

For a spine there are one of these in every vertebra. You would use this to take an orientation from a constraint made with a few nulls and blends in a subnet and applying it to the bone.

I was spoilt in the past with my Celeron 500 with animation master and having 218 bones in a rig with maybe 300 constraints in the rig and having it as snappy as hell no slow down at all! It had all Quaternion rotations so all rotation constraints were predictable and no worrying about gamble lock ever. But animation master is only a cheep program so it can

[Jason]

I just want to say that I think you're doing some good work there, meshsmooth. I'd love to see a complex rig perform in Houdini.

[meshsmooth]

Well if I pull my finger out it shouldn

[TheUsualAlex]

hey,

If you guys have any ideas of how you want a generic auto-rig and what you want it to be able to do, can you please post your ideas here. If you can post a sample of your hip file too that would be awesome.

I started to write some generic autorig hscript about a year ago, but never finished it because I was finally busy. I'd like to start looking into OTLs these days.

Cheers,

Alex

edit: Oh, I think you can post the maya files too. Houdini would be most preferred.

[edward]

Why not use a Path object instead of doing a rotation constraint per spine bone? A lot of this workflow was based on trying to simplify spine rigs in the first place as Jason Iverson and I were discussing about this.

[meshsmooth]

I don

[HarryH]

How about using a blocking spine which has same parms as the final spine with the strech, squash and twist controls. This can be swaped in by animators after the initial blocking of the animation with the lighter spine Rig?

-- Z

[meshsmooth]

The shoulders need to end up at the exact same position at the end of the chain Otherwise the head will not be in the rite position ether and if the arms are in FK mode the inconsistency will be amplified.

The spine I am currently dealing with has 7 vertebrae and gets its overall shape from 3 bones.

To blend 3 orientations into one output it takes between 10 and 15 nodes. This is because blending more than 2 orientations in Houdini with 1 blend object has to be done in Euler blending that is unpredictable and quite frankly is a JOKE!

To blend 3 orientations and apply them to 7 bones I need 51 nodes. Mind you if the blend object blended rotations with quaternion interpolation I could have 20 less nodes and if the blend was incorporated into the bone object (with quaternion interpolation). I could have it work in 7 bones and there would be on need to even wrap it up in a subnet.

Houdini only needs to make a few changes to make its rigging up there with Animation Masters and that is miles ahead of Maya.

[edward]

What in particular don't you like about how a spine based on "bones to a curve" deforms? As for speed, it sound like it'll be a lot faster than doing 51 nodes. If you create a 2 CV path object, that it gives exactly the same controls as a 3 bone spine driver.

How would you expect to blend 3 inputs using quaternions? Do you blend the first two, take that result and then blend with the third? Note then that if you re-ordered the inputs, the result will be different even if you moved the weights along with the inputs. I'd be extremely interested in seeing how you blend 3 orientations with between 10 and 15 nodes. Could you post a file?

[meshsmooth]

I developed and tested the solution in Animation Master. Where all rotations are calculated in quaternions by default. I made 4 bones pointing in different directions and then made one other bone that had 4 orientation constraints on it. This was what I saw as the correct solution to the problem. Any solution I could make in animation master using only orientation constraints that blend between only 2 objects at a time, position constraints or aiming, would be able to be pulled off in Houdini. I tried the 2 orientations blended together and the other 2 orientations blended together, then there outputs blended. This didn

orentate_like_3.zip

[edward]

Hey meshsmooth, could you give a description of how you're blending the 3 orientations? It's a bit tad confusing just looking at the hip file.

[meshsmooth]

Err look at the post above.

i will describe it differently if you need me to, but rite now i am procrastinating about work i need to finish today. Maybe in a culpa days

The important you may or may not get that the final blend is doing a look-at. Did you know they can do that? Orient axes is enabled when you have 3 inputs, tick it and turn off the transform rotate and scale on input 2 and 3 and it is a look at with picked up vector. This is where the quat workarounds that I have made come together.

I think there should be a new VEX / VOPs context that can take the objects transform from one or more (if you define it, just like a vex sop) and you manipulate it with vex or VOPs and export a transform matrix TRANSOPs. I can here you saying rite now but we already have chops and Vex chops. I say to you vex chops are so limited you can

[calin_casian]

If you guys have any ideas of how you want a generic auto-rig and what you want it to be able to do, can you please post your ideas here. If you can post a sample of your hip file too that would be awesome.

I'm working on such "things" for SESI's SIGGRAPH character demo. Thay will be availabe for download after SIGGRAPH.