Dynamesh - What a dumb f***k

[MikeANash]

Hopefully the title got your attention hehe

Just wondering if anyone has a advanced understanding how dynamesh works exactly? Finding it rather stupid at times when you need that extra polycount or it could be just me hehe.

Dynamesh seems to be sticking to certain polycount when working with certain meshes, I have some meshes that sit around 1mil while others are on 2mil with even less details.
I've tried re-project and re-importing with no luck,
also tried up scaling .

I guess i'm just wondering what im doing wrong or is there a hacky way to increase that polycount?

Here's couple examples of whats happening.

I tried a little test, very basic, odd result.
Both 1024 res, Lower polycount resulted when doing remesh, tried with reproject to get the same result.
9109

Example of one mesh being higher then other when I would need the arm to hold a higher polycount for further work.9110

Any inputs??

[Kel Solaar]

Hi Mike,

The dynamesh is most likely a voxel cube fitting around your model bounding box, now imagine that you have a very elongated model(Like snake or something like that), at the same resolution (1024 cells for example) than a more compact mesh (A head), your cube cells will resolve far less detail.
The closer your object is from a cube shape the better resolution you will be able to resolve, when you start to scale the voxel cube by elongating the mesh you are losing resolution on that mesh.

I attached a small image to represent the thing, see how the arm has less voxels to represent its volume compared to the hand. I quickly counted, the arm is using around 26 voxels here whereas the hand has nearly the full cube with 56 voxels (More than 2 time the resolution in that 2 dimensions representation).

I did my first tests doing kind of snake and got the same troubles than you. The only way they can handle that problem is by doing an adaptive voxelization providing more voxels where the detail reside (It can be compared to Renderman BrickMaps for that matter or any similar data structure), I'm sure they have that kind of working but it must have its drawbacks (Speed certainly ) Hope it makes sense!

KS

[MikeANash]

Thanks a ton for the explanation Thomas! I guess we are still bounded by thinking within the box hehe. I had no idea it worked like that, good to know.

[Kel Solaar]

You are welcome :)

Though I can't claim at 100% it's working like this because I didn't saw the underlying code :)

KS

[BrettSinclair]

All i can say here is.... o_0.

[MikeANash]

Kel Solaar - I think your right on the money Thomas hehe. Simple test shows proof to your Techy understanding of things :P
9112

BrettSinclair - Haha

[Kel Solaar]

Mike: Theory proven by experimentation :) Cheers!

Brett: Hehe :)

KS

[Locusta]

Wow guys. Big thanks. Another mystery solved. Was wondering how that works. Well, let´s see what ZB 5 brings. In the meanwhile we have to make our work-arounds like always LOL

[Storm]

Hey Mike, just in case you didn't know, you can freeze the SubD as well, that way you might not have to work with such high polycounts when using dynamesh. Just in case you weren't aware.

[MikeANash]

Storm - Yeah, I might have to result to using that, just kinda lags the workflow down.

[Crispy4004]

I'm only finding it practical as a base mesh building solution. I was really hoping Dynamesh would be like an improved Sculptris equivalent, supporting more polygons with auto density management. I wasn't expecting what is essentially a re-tooled ReMesh. It still seems useful, but it doesn't quite meet my hopes.

[Ballo]

Is a good start to do fast sketches and get the freedom that you get in paper. Although the old method, build your basemesh, you get clean results, old school. :D

[hbajramovic]

Had the same problem as Nash. Thanks guys !!! clap