edward

You use a MDD ROP (Output) node to write out MDD caches and you can read them back in using MDD SOP (Geometry) nodes. If you want instancing of the sequence, I think Packed Disk Sequence primitives are what you're looking for http://www.sidefx.com/docs/houdini/model/packed.html#packeddiskseq

I think the easiest way is probably hit Ctrl+A in the network editor to select all the objects. This should scope all the translate/rotates on them. Now you can just scale them in the Animation Editor > Dopesheet.

Can you save the output of the Agent SOP to a .bgeo file and post it here?

Why do you need to run it with sudo? The steps should just be something like: $ cd /opt/hfs16.5.XYZ $ source houdini_setup The Houdini 16.5.XYZ environment has been initialized. $ cd ~/Documents/wherever_your_bvh_is $ mcbiovision Usage: mcbiovision [-e -b -d] [-c] [-v] [-j object_script] [-o -w -f -s -i [-n -t -l chainfile]] infile [outfile] This program converts a BioVision motion file to a Houdini script & channel file(s). Run the script to create the geometry and channels. ...etc...

What happens if you render the agent in Mantra?

VDB Visualize Tree SOP or perhaps Volume Slice SOP depending on what you need.

Just the heads or the entire agent? For the entire agent, probably a POP Look At. For just the heads, see http://www.sidefx.com/docs/houdini/crowds/states#lookat

When it better to uses Volumes over VDBs? - When you don't have as mentioned above - When you have a data on nearly every voxel, making it not sparse anymore.

PolyFill is one of the new options in H16 to cap holes. It's worth a try.

Is your .hip file logged with steps to reproduce? We have lots of code to deal with these situations. Furthermore, I've created set ups where I've had networks loop back on itself as part of its normal operation.

I spoke with one of the paper's author and he mentioned that, "The implementation of the underlying MPM solver was done in OpenVDB and all the meshing and renderings where done with Houdini (using the native VDB nodes)."

If you mirror all the IK bones using the object viewport's Mirror tool, it should copy all the IK solver and goals for you.

Hand-keyed is probably still "audio driven animation" albeit with a lot of manual work. The other aspect is that there's nothing wrong with manual tweaking on top of a procedural solution as well.

How would $FD / $FFD behave differently to solve the issues?

The traditional resource for this is the e-book "Simultaneous Music, Animation, and Sound with Houdini" by Andrew Lowell. However, his site is now down so I don't know if it's still available for purchase. However, I think Peter Quint's tutorials are still available: https://www.sidefx.com/tutorials/chops-and-music-driven-animation-i/ https://www.sidefx.com/tutorials/chops-and-music-driven-animation-ii/ https://www.sidefx.com/tutorials/chops-and-music-driven-animation-iii/