blending between different topology curves

[pasto]

Hello,

 

I would like to animate a smooth morphing effect between two letters, two set of curves then, without the same topology or even point number.

I tried to interpolate a point number value to make the points travel to their most logical neighbour to prevent some huge twist during the animation. I created a num attribute corresponding to the ptnum of the destination curve, interpolated it using a xyz VOP, and used it as a blend shape or sequence blend attribute, but obviously the interpolation is position based and even with a serious resampling I can't reach the details of the shape by just selecting the neighbor points. (from one curve to the other)

Here is a file with my results so far, I guess I am missing something.

Is there a way to resample two curves with the same number of points ?

Should I use the curveu attribute of the destination curve ?

 

 

Many thanx.

Jérôme

 

 

 

 

 

 

ferire.hiplc

Edited January 19, 2017 by pasto

[LaidlawFX]

Your scene is borked. The locked node doesn't show the geo. Probably convert it to geo first, then lock it at a null as opposed to an alembic sop or file sop.

In general for morphing if you handle it like a flip sim you should be good. Convert your end states to a VDB, and make sure you hit the option to fill the shape.

For the blend Make a point cloud from both the shapes and then blend sop between the shapes. Put it through some noise for variety so it's not linear. 

Convert the particles to vdb, and use the VDB Reshape and use the close option. 

If you are truly doing curves, then you'll just have to lay them out cleanly. Or sort them in Houdini so the point order is correct for the blend, and the resample based on a number of points so the number is correct. Place it through a noise field during the blend for some variety for the linear blend, or make a few in between states.

 

[pasto]

Sorry, I updated the file.

Thanx a lot Ben, I will try all this.

So having different point number is absolutely forbidden ? I htought I could send several points to the same location.

 

 

 

[LaidlawFX]

if you use an attribute called id with the blend shapes or sequence blend sop then yes

Edit: Ahh, this is what your num attribute is doing.

Edited January 20, 2017 by LaidlawFX
num = id

[LaidlawFX]

I see the curvature issues you are having at bases of the m and such... I don't have a good solution to 100% toolify it, someone else might. 

The best hack I keep coming back to would be to pre-warp the initial shape with an edit sop to the end shape prior to your xyz detection, this maintain using your same method. It's hell of dirty, but if your only doing say a few words or just your name it be quicker than solving the more complex issue of a generalized tool. Maybe an hour at most. Your blends would be clean. If you had to do a massive paragraph or a whole credit roll, I would custom make an alphabet to handle the special cases in it, and then make a tool that swaps out each letter that doesn't work with the general method. 

[Yader]

Attached is a possible way of blending different point counts with POPs and Pop Attract, but beware, this won't work properly with all fonts/glyphs. Sometimes you'll get artefacts. Best way to approach this is to resample, so the letters have same point counts.

Pop_Attract_DifferentPointCount_01.hipnc

[Atom]

What a fun scene to play with.

 

It also seems like the second word needs to have the same character count as the first input, even if you resample to make each side have the same number of points.

 

[pasto]

Thanx a lot guys,

Ben, my aim was to make a somehow generic tool, strangely the Illustrator blend is doing a better job at blending when the starting point is well positioned.

Yader, for this sort of (dancing!) behavior the blend shape makes a pretty good result, do you think we can resample a curve at a fixed number of points ? 

I also tried to orient the normals to the corresponding point on the target curve (pretty much the closer one but this method never grasps the whole shape details).

Many thanx

 

Jérôme