Geo-Scatter - 5.4 Scatter Groups

[BD3D (Dorian B.)]

@ChameleonScales

So I found a way by adding a raycast node (+Object info and set position) in the Scatter Node tree right after points are distributed. There’s more digging to do if you want some features to work, like have the instances be aligned with the terrain’s normals.

The main problem with this technique is:

1. orientation:, let say you align your instance on the Local Y, what is supposed to be the local Y axis now that we have projected the points?
2. attributes, how are we supposed to paint vertex groups/vertex colors ?

We could implement it, but that means this mode will have less options, because we’ll hit a geometry node limitations wall Not sure how familiar you are with geometry node, the blender Devs spoke about one major limitation we keep facing, earlier this week. the lose all attribute problem
https://blender.chat/channel/geometry-nodes/thread/mvc3XeG43rvfXxNNa?jump=MbWJNTH8hNu7ssFDh
https://blender.chat/channel/everything-nodes?msg=dP75o3HCuqFF6a4eB

[ChameleonScales]

I see. I think it’s ok if some options aren’t available. Right now I’m using my trick and the difference in performance goes clearly from several seconds/frame to several frames/second.
For the orientation on the face normals, I think it should be possible by going inside s_distribution_random. I’ll try to find a way.

Even if it’s not possible that way, the emitter mesh can always be projected on the terrain to match its face normals closely enough, and if more precision is needed, it can be subdivided. It would still be way simpler than the terrain itself.

To give you an idea of what I’m working with right now, my terrain has 4 million faces (2 by 2 kilometers with a step of 1 meter).

[BD3D (Dorian B.)]

I marked this in the idea list

Note that the vertex groups, vertex colors, material culling mask, and the new visibility select features are all optimized when working on such conditions. We always deduce masks from the density when we can

[ChameleonScales]

Thanks!
And sure, all those features help a lot already. Well, except when creating the system, since you don’t have access to them at that moment. BTW isn’t there a technique to avoid the long loading when creating a biome on a large mesh ?

[BD3D (Dorian B.)]

@ChameleonScales

Well, except when creating the system, since you don’t have access to them at that moment

You do, in the operator settings menu, a popover right next to the operator, also available in the biome header

[ChameleonScales]

Sorry I missed that. I didn’t find it in the documentation either. Maybe it could be mentioned in this section?
I just tried using these options (on a 16 Million faces 8 by 8 kilometers terrain) and it took ~70% of the time it took without using them. So yay!

[jcfnav]

Hello,

Would it be possible to expand the usability of Culling Mask’s Bezier-area a little bit further?

When receiving blueprints from clients (usually dxf or dwg) the grass areas are usually ready splines so it is easer to define the grassy areas without doing any extra steps.
Now, the idea is to be able to have the same controls that we have in the Cull Near Object where we can “bleed out” the grass from the grassy area.

Currently I have to create a mesh object from that spline and offset it outwards and then use that object in a separated ocllection in the Cull Near Object , too much hassle if the Culling Mask’s Bezier-area can be developed a little bit further.

Hope it makes sense

[ChameleonScales]

… That said, all these optimization features still don’t come close to the “smaller emitter” technique I’m using. Just creating the system takes way less time on a smaller emitter, regardless of whether I enable optimization options and disable visibility before creation.

To give you more info: what I did was establish an area from the camera where grass will be emitted, since beyond a certain distance I just use a texture on the terrain surface (and of course I also use Scatter5’s Camera optimization options). That emitter surface goes ~300 meters away from the camera (can be more or less), and of course doesn’t go behind the camera. Compared to a 2 by 2 km area all around the camera with millions of faces, the difference in performance is I imagine somewhere between 1 and 2 orders of magnitude.

[D_F]

Hi Dorian,

Just wondering if there is a simple thing that I am missing regarding switching to manual mode. Hopefully it is easy to pick up what I have missed. I am trying to dot some trees on a ground surface.

Blender 3.2
Scatter 5.2

• I get my large surface prepped (undulating hills going from the camera to the background).
• I select scatter and apply it to that surface.
• I apply the Forest 7 preset (large trees and shrubs), which populates it nicely everywhere.
• What I really need is a line of trees and a few accent trees in the foreground, so under the distribution tab I switch to manual mode, and enter manual mode.
• I select the dot placement tool from the viewport left menu.
• I dot some points on the surface, but there are no trees.

I would have throught the steps above should have allowed me to click and place. Coming from a decade of using Forest Pro, I’ve tried a number of the more obvious things but seem to be loosing time to it so any help would be appreciated in troubleshooting this.

Thanks,
Daniel

[BD3D (Dorian B.)]

Hello Everyone

@jcfnav

Now, the idea is to be able to have the same controls that we have in the Cull Near Object where we can “bleed out” the grass from the grassy area.

Like a “blur” option? it’s already on the TODO list
however this kind of contour control is something we rather implement

Masking 101 in Blender (2.93) | Part 1 - Overview (NOT a tutorial)

We already have a working prototype, unfortunately it got a bit buggy at intersection so we need to do more research

@ChameleonScales

To give you more info: what I did was establish an area from the camera where grass will be emitted, since beyond a certain distance I just use a texture on the terrain surface (and of course I also use Scatter5’s Camera optimization options). That emitter surface goes ~300 meters away from the camera (can be more or less), and of course doesn’t go behind the camera. Compared to a 2 by 2 km area all around the camera with millions of faces, the difference in performance is I imagine somewhere between 1 and 2 orders of magnitude.

Thanks for your input, we’ll do some research over the subject

@D_F

I would have throught the steps above should have allowed me to click and place. Coming from a decade of using Forest Pro, I’ve tried a number of the more obvious things but seem to be loosing time to it so any help would be appreciated in troubleshooting this.

Hi Daniel,
Welcome to Blender Artist

I believe Fp propose a “edit” button, manual mode is not an edit button, it’s a distribution method on it’s own. For the time being the distribution can be either fully procedural or fully manual distribution, there’s no “hybrid” between the two.

In consequence the workflow is a bit different, in order to achieve your goal, you need to create two system(s) instead of one:

1. create a new procedural distribution layer with your forest preset
2. create a new manual distribution layer for your details, note that in order to use manual mode efficiently you can use the “Manual Scatter” operator instead

I believe that you did not see new instances probably because a procedural feature is interacting with your manual distribution, unlike forest pack, our procedural features can still interact with all manually placed points, in your case i believe the procedural texture of the forest culled your points

If you have any further question, suggestions, or share the potential frustration you encounter when switching to blender & our toolset, please let us know!

[patchar]

Hello guys

First time here. I’m considering purchasing an addon but I have some question.
Is it any possible way to export file with scatter addons or convert the animation to render on renderfarm.
I’m working on animation project which require movement on grass from wind and I’m really sure that my laptop can’t handle all of the frame I need to render.

Thank you so much

Patchar

[BD3D (Dorian B.)]

Hi @patchar,

Our export functions will not support any animation from our wind or offset features.
However, If your render service accepts .blend files, then there is no need to export anything, just make sure to disable the camera optimization features as this feature relies on our plugin scripts. Everything else should be fine as it is part of our geometry-node scattering engine.

I suppose you’re the same person who contacted us this morning on the BM?

PS: if your laptop is not that powerful we strongly suggest that you take a look at our optimization guide

Documentation - Optimization

Cheers
Dorian B, founder & lead developer

[patchar]

Thank you so much!
I’m purchasing it now.

[BD3D (Dorian B.)]

Thanks for supporting our work,
Don’t hesitate to contact us if you have problems or feedback

[patchar]

Hi

I’m working on a large landscape scene and having a problem.
Filling large mountain with only grass is such a pain. Is there any LOD option for scatter asset to render faster. Any suggestion?

Or Should I use low detail grass to scatter far object the only asset in the front?

Thanks

[BD3D (Dorian B.)]

Hello

@patchar

Filling large mountain with only grass is such a pain. Is there any LOD option for scatter asset to render faster. Any suggestion?

Indeed i could give suggestion depending on your scene
Could you give us a general idea of your scene? Is it an animation, or a still shot? Do you need multiple views or is it a single shot artwork? a screenshot would help.

I suggest taking a look at our optimization article where we explain everything to know about the limitations of blender and how to overcome them

Documentation - Optimization

[Eric_Giordano]

Hi Dorian, I’ve been using Scatter 5 and it’s awesome but I have a few issues/questions. (Also I just downloaded Blender 3.2 FYI)

1. A few issues with Manual Scatter
   a. Some of my manual scatters I made a few weeks ago aren’t showing up any more when I try to display them
   b. When I do a manual scatter and use the Lasso Fill, sometimes I don’t see the bounding line as I draw an area to scatter into, even though once I stop drawing the area, the scatter works.
   c. Also when I use the Lasso Fill, sometimes the scatter doesn’t fill the entire area I select, only a small portion.
   d. Also with Lasso Fill, the scattered elements show up on the top AND the underside of the emitter. I want them only on the top surface I drew onto. (Also when I tested drawing on a cube, it scattered on the side I drew onto and ALSO onto the “back side” of the cube, not what I intended.
   e. Also with Lasso Fill, Sometimes I get an error:

        Traceback (most recent call last):
                File "C:\Users\....\Scatter5\manual\brushes.py", line 883 , in modal r=self_modal(context, event
               File "C:\Users\....\Scatter5\manual\brushes.py", line 771...
               File "C:\Users\....\Scatter5\manual\brushes.py", line 9781...
               File "C:\Users\....\Scatter5\manual\brushes.py", line 9419...
        ValueError: connot convert float to NaN integer
   

   f. I’m wondering with manual scatter, I would love to be able to tweak the distribution AFTER I manual scatter, but it seems I can’t do this. I can only figure out how to set the points per interval BEFORE I scatter. Is there a way to do this?

[BD3D (Dorian B.)]

Hi @Eric_Giordano
Are you aware of the compatibilities information?
https://sites.google.com/view/scatter5docs/changelogs?authuser=0

For blender 3.2 you will need Scatter5.2

a. Some of my manual scatters I made a few weeks ago aren’t showing up any more when I try to display them

Yes that’s because you need to upgrade to 5.2 and click on “update node tree”
Please let us know if you still have these issues after you upgraded to the appropriate version

[ChameleonScales]

Proposal :

I suggest replacing this:

by this:

Some possible icon sets:

I don’t like choice 3 because the screen seems to indicate all viewport modes which is misleading.
I’m partial between choices 1 and 2 because choice 2 is visually more contrasted than 1 but also a long list of flat circles may look like big radio buttons .

What do you think of all this?

[BD3D (Dorian B.)]

@ChameleonScales

I suggest replacing this:

Hi thanks for your proposal, very appreciated!

“Un-necessary dupplication”

Actually, we just got rid of the double panel structure in 5.3
This is what the UI currently looks with the “dark box” UI option disabled

Note that the features indentation is an important part of the interface look, we gave a shot to a panel with an indentation structure, just like native blender, following their GUI rules, and it looked quite odd and very messy in the N panel, this kind of UI structure only works on the properties editor IMO

That said, we still consider that opening/closing per predominant features might be a good idea, we just aren’t sure how to implement it

We tried this, which might take too much space in the interface

or this,

and there’s a BUT
not not sure how the arrow would be elegant in combination with our icon GUI option (yes we have this option for the more visual users)

“Separate categories for similar features”

The viewport method options only makes sense for the features designed to optimize the viewport,
while i agree that “face select” could also be in the culling mask, i hardly see how a vertex group could be useful for viewport viz

The whole mask-culling category is a little strange tho, because at some point we could introduce falloff distance to let’s say a bezier area mask, which could also have an influence on the scale of the points

also one of the specificity of masks is that you might want additive or substractive ones, by changing order ect… which is basically a modifier stack and we’d like to avoid this logic. If users want the ultimate flexibility he should jump in the nodetree right? assuming the nodetree is easy to understand which is why we are currently doing some large cleanup in our geonode engine noodles

“Viewport options dispatched”

It is needed, because you might want some optimization features only for the solid viewport (ex display) some everywhere (like camera culling) and some only for the viewport (ex face preview)