Future Particle « Hair » Nodes discussion, and everything wrong with « Hair » particles

Hello.

I am and environement and archviz artist and the author of the « Scatter » addon and I have a lot to share for everything that concerns vegetation scattering in blender.

So Can we talk about the future particle « hair » here ? @jacqueslucke
or well anything related to the future of scattering within blender.
I don’t know if hairs are going to be separated from emitter after all, if there’s going to be a new particles hair node ? if everything will be mixed in one node editor? not clear.

We should have a dedicated period of development for Scattering.

Scattering is how people use the particle system the most within blender and there’s a clear lack of features in this area. for many years. users keep asking for improvements and nothing is done. That’s why i made Scatter, it’s an attempt at re-thinking the current workflow and bringing new improvised features in python, that should really be done with C at the first place.

I have no clue about hairs and furs, not my domain. the same goes for particle node and all things motion designs. not my domain either, and way too complex for just scattering things around.

Therefore All my post below will be Scattering things concerned.
In the post below i’ll share thoughs about flaws, ideas, improvements
i’ll try to restructure everything when i found most of the important ideas.

4 Likes

Well, I’m not sure about hair yet, but as for scattering I know that the team have remarked that the array modifier, and indeed any other modifier that instances geometry is inefficient and duplifaces are preferred. I’m pretty sure in the future, with everything nodes, we will get more options. In specific I’m hoping for something similar too Houdini’s approach with the “Scatter Node” and the “Copy to Points” node.

Straight up copying to vector points is very efficient!

2 Likes

First flaw:
there’s no influence sliders for vgroup.

if there was a simple scale influence thoses annoying baby trees wouldn’t exist.

Annotation%202019-09-21%20115941

exactly like the influence value in the texture properties
in textures but for vgroup.

3 Likes

Could you explain in further detail. I feel your current explanation is still a bit ambiguous.

Flaw 2

No way to keep a particle sys density consistent through its usage.

having a particles count per m² or km² could be really more useful, as it give a true idea of the density.

Flaw 3

If not enough display option, visualizing complex biomes is almost impossible. Or at least not confortable at all. You cannot expect blender users to play with psys parameters blindly with bounding boxes.

so here’s contents that need to be implemented:

  • “display as proxy” in > properties editor > object > viewport display options
  • “display as point cloud” in > properties editor > object > viewport display options
  • “display as decimated” in > properties editor > object > viewport display options

please see post above for more explanations

3 Likes

Fourth idea:

Support for camera clipping and distance culling within the particle system.

for sure we could use a modifiers do to so (like i did in the addon) but this feature is so important we should have an native support for camera clipping and distance culling within each particle system. a simple toggle camera clip + threshold options, and distance culling toggle + distance options would do the trick easily.

4 Likes

Flaw 5:

there’s no way to use cycles powerful node for particles distribution. It’s a huge flaw because cycles dispose of very powerful tools to make such distribution procedural maps that the texure editor simply don’t have. Also, in most cases, particle distribution is directly linked to the terrain material. For example certain colors in the terrain material will never have particles (like rocks, snow, water, ect)

maybe a new modifier “material to weightmap” could do the job
-> from material data to bitmap (i think there already a shader for that) -> bitmap to vweight.

Flaw 6:

there’s no way to make the particles not “intersect” with each others. either within the same particle system, or across multiples part sys. It’s really problematic. in most cases, you don’t want a scattered assets being inside another one.

5 Likes

But blender have their culling system, maybe not in cycles

1 Like

But blender have their culling system, maybe not in cycles

could you explain ?

Blender uses automatically a camera culling in viewport. For that reason performance changes depending where you see

Well why my culling technique give much more performance then if blender already do that ?

Maybe because distance density culling and because your system don’t calculate culling by each frame

Flaw 7

The Weight proximity modifier is really important, using it allow us to do “boolean” operation on the particles.

the current problem with this “Weight proximity” modifier is that we can only do operation with one vgroup at the time.

so if we want X objects to substact weight from let’s say 5 particles system, 5 modifiers will be needed (if we want to keey having different vgroup influence per psys of course, and it is really important to do so)

having a modifier that can substract multiples vgroup at the time could be much much more useful and maybe performance friendly

of couse, another more easy solution could be implemented for users than this modifier. ( a node form ?)

3 Likes

Flaw 8

with random rotation we can randomize the particles on the local xy.

with random rotation phase we can randomize the particles on the local z.

with random scale we can randomize the scale of the particles on local xyz.

but why not having it vector Based ?

And why random phase is from 0 to 2 ?
How can something be 200% random ? 100% is more than enough. It should be from 0 to 1.

rotation randomisation
-x
-y
-z
scale randomisation
-x
-y
-z

this seems more logical and give more possibilities over its randomness. right now it’s kind of a bit confusing :slight_smile: “phase” being z, “rotation” being xy.

also another things that is confusing is that there too option for controlling the scale of the particle.
either with the scale ratio or the hair length. only one is necessery, and it should be a 1:1 ratio by default.

Flaw 9

there’s a texture mixing calculation type avaible in the code but not avaible for users ?

3 Likes

Because it’s something completely different.
I guess @Alberto is talking of frustum culling. That way geometry that lie outside of the cameras viewing frustum get culled before rendering. You have to think of what you are saving when. Anything that gets done before rendering itself can cause a lot of workload and won’t get better by frustum culling. If you are processing hundrets of thousands of particle positions you can’t cull anything here before the position and/or dimension is final and the rendering doesn’t really have to be the bottleneck, but the computations before. Another thing here is if you put less elements into a scene with that camera check you implemented and then blenders culling skips anything outside the camera, you are saving multiple times, because you calculated less elements, you generated less and by changing the camera view even less remains on screen.

2 Likes

Flaw 10

maybe a bugs ? -> representation of particles inacurate in the viewport

first case:

  • the rotation is set to random and count display on 1%
    but somehow at 1% all the particles rotation are not random. so using the display % is actually a bad way of previewing what will happend with particles distribution.

  • second case: having two textures influencing density/length will make the particles representation in the viewport totally inacurate when chaging the values. an edit mode refresh is needed to make them precise.

Flaw 11

PERFORMANCE ISSUE of course

  1. lot of particle’ instances slow down blender. other sofwate can handle instances much much better at a much much higher count !
  2. it seems that the outline overlay is drastically slowing down blender performance on high particle count. something wrong with this design, maybe ignore particles with the outline.

Flaw 12

No float number support for particle display % ?
when working on really dense grass field, 1% isn’t enough.
the ability to go below 1% is necessary for a lot of cases.

Flaw 13

huge problem for animation, any modifies r that influence particles density vgroup will change the seed of the particles. i tried this with every >Emission>source> distributions options. nothing works. particles seed change every time.

this is just a simple weightpaint set up.

we should have an option for hiding particles influenced by a vgroup, instead of regrouping all particles in the painted area.

1 Like

Flaw 14

particle seed in viewport is unpredictible.
Sometimes, when changing values in the particles parameters, textures or modifiers, the displayed particles in the viewport are not in the same place they will be in the final render.

this is a big deal for important object on your scenes like trees and big foliage.

i couldn’t identify the cause of that. the “viewport seed” just behave strangely

the only counter to this annoying ‘bug’ (?) is the user constantly toggling edit mode on/off or toggle viewport visibility on/off

Idea:

in my addon i propose safety features:

if the obj scattered is too high (after X polys) the addon will automatically set the particles to a bouning box

if the polycount is too high, the display % will automatically adjust accordingly to not go after Y displayed in viewport

if the polycount is wayyy too high, the addon will automatically hide the particle system from display.

thoses features could prevent blender for crashing to a sh-ton of peoples

of course we can take a look to clarisse ifx

this program is just magic, it can show off millions of object, billions of particles live in the rendered viewport

i don’t understand how they can achieve this
having blender 2.8x viewport being a lot more optimized for instances and particles is a must for any environement artists who’d like or who are using blender.