Considering that you can do new things with this approach that were impossible before, there are things which are now infinitely easier.
So, by your own logic 4.1 is infinitely better, no?
Do you really instance as much, before finishing your modeling session ?
If you have to modify an instanced mesh, you will probably go into local view to change it.
During that time, using modifier will not be a problem.
And when your modeling session will be finished, you just have to apply it to make auto smoothing change permanent for all instances.
That may be annoying if seeing all of them matters for modeling session, and generally, in those cases, you only need the few of them, at bounds of original mesh.
The past couple of posts had nothing to do with GN⌠youâre in no position to be talking about commitment. lol.
âJust change your workflow lol, why are you even working in that way?â Is never an answer to regression. Terrible attitude by some people here.
2 Likes
That is not my reply.
A correct rephrasing of my reply would be :
âIs what you have shown in your video a real practical case ?â
Havenât tried this before. But it looks like a hassle to do in a more complex scenes with multiple collections. Still doesnât solve the problem of having to click every time you exit edit mode.
I assume youâre talking about geometry nodes. Not everyone knows and uses geometry nodes like you. So to them these possibilities are close to useless.
I believe being in local view still have performance impact when thereâs many instances, especially with modifiers. and you cant apply a modifier to an instance and have it still be an instance.
1 Like
Itâs not your place to question that. There are many people with many different workflows, I indeed like to edit mesh after instancing. Iâm doing shape key animations, I have instanced objects, I need auto smooth on all of them, shape keys can sometimes change edge angles, I canât work with new way with modifiers, and Iâm certainly not going to remember to click a button every single time I use grab tool on my mesh.
Youâre right. Unless instances are excluded from the viewlayer performance still takes a hit. And again, you shouldnât have to do that much extra steps for something that was automatic and fast too version earlier.
2 Likes
Tools will become available by ppl who do know GN, yes⌠and everyone will benefit from that.
Having the Auto-Smooth applied when you exit edit-mode is something which you can expect a python script for as stop-gap until something along those lines make it into Blender. In the meantime - no-one is forcing you to use 4.1.
Good luck!
1 Like
On a regular basis, yeah. All the freaking time.
1 Like
Heh, this is exactly the problem I predicted to happen, I just didnât have the time to verify it would. And I didnât expect itâd be so severe 
2 Likes
There are workarounds thoughâŚ
You can have a realized instance that you can duplicate instead or you can use collection instances.
If youâre regularly doing 14,000 instances like in Jaxxâs example by manually pressing alt-d then you need to re-evaluate your workflow.
@rawalanche - it is only severe if you willfully choose to stubbornly refuse to use alternative methods for instancing that have been around since 2.9⌠Otherwise it is a non-issue.
1 Like
You are right. I forgot that change. People complained about the fact that applying a modifier was modifying instances.
But you can apply modifier, anyways, to a new object data, and then relink mesh data of instances to new datablock.
But if there is no other modifier, you can convert to mesh to have a quick update of instances. But that cannot work for characters with an armature modifier.
If your mesh is created with its shapekeys and mark as an asset in its own blend file, and then linked to be animated with library overrides in another one.
In theory, there should be no problem related to set-up of modifier.
It should be there when object is linked.
In practice, if the use of modifier has a performance impact after changing object into an animated proxy; I agree that is a regression.
If you have a lot of proxies with different animations and a big regression ; that is a not negligible issue.
But that is not the same case, as if you want same animated object, repeated a lot of time in scene with same animation. For that, passing through collection instances or GN instances, instead of Alt D instances, should not damage performance.
But 4.0 is not release, yet. It is possible that devs manage to solve performance regression before 4.1 release.
I am reading none of that just so you know. In 4.0 itâs single button. Iâm not doing paragraphsâ worth of work for 4.1 because devs broke something. Iâm simply not gonna use any more Blender versions if that remains the case. I have a job, I have a workflow, if Blender no longer fits that weâll reevaluate our options.
How about instead you mind your business and stop telling people what workflows they should use? Some people have actual productions that create actual products instead of cute GN renders.
If you canât find a way to continue the discussion without questioning or disregarding other peopleâs workflows, or being condescending, you should stop contributing to this topic.
5 Likes
Wow⌠youâre saying that without a hint of irony. Maybe take your own advice.
Thanks.
2 Likes
Keep in mind that Blender 4.1 is still in Bcon1 and there is a lot of time for the devs. to iron out regressions in workflow. This by no means is a last minute addition to version 4.0.
Eventually, we will need a rethink on how the modifier stack works and make it far easier to work with when many objects need editing at once, much like how the creation of node trees and node groups got upgrades when it started to become a more central part of scene building.
9 Likes
Sure⌠thatâs why we need to discuss that now. Never think that others will do the thinking for you⌠HUGE mistake.
Autosmooth in my experience is annoying af.
My workflow is setting autosmooth to 180, meaning I will control my normals myself.
Lets say I model a midpoly asset (bevels, mid-level topology) that is separated into 100 objects with different modifier stack.
Weighted normals with smoothing groups, also called Face Influence, control the shading. I can easily assign Weaker influence to bevels by ticking âFace Influence: Affectedâ within Bevel modifier which makes this workflow procedural with Weighted Normals âFace Influenceâ tick and Threshold 10
Now everything looks perfectly shaded - all flat surfaces have flat shading and I apply all modifiers. Still all separate objects have perfect shading.
I apply all transforms. Still all separate objects have perfect shading.
I now have 100 objects that have custom normals on all of them.
I join 100 objects together, which shouldnât change anything⌠voila, shading breaks.
How this can be fixed, so this doesnt break? All objects had to have same Autosmooth settings before applying modifiers. Huh?
Havent tested the new shading system but hell, autosmooth is cancer. In my opinion.
4 Likes
I mostly use a mid poly weighted workflow for gamedev, so was initially concerned that the new changes may screw with old files. The good news is that old files will be fine. Just a matter of deleting the geonodes smooth by angle modifier that is added, as it is no longer required.
To really understand the new workflow, it is important to look at the attributes panel with regards to the sharp edge and sharp face attributes. By default a primitve will have the sharp face attribute added with all faces marked on, to give a flat shaded look. If you delete this sharp face attribute, it is the same as using the smooth shade operator.
For the weighted normal workflow in 4.1 you donât want the sharp face attribute, so either delete it or right click and shade smooth. Then it just a matter of adding the weighted normals modifier and tagging sharp edges and selecting strong/weak face weighting as is required.
As can be seen in this image, I imported an old model of mine and there is no need for anything other than the modifier, sharp edges and the weighted faces attributes.
(note: it was suggested that I post this here, as I originally had it in the meeting notes)
12 Likes
Beautiful seeing proper normals setup and management in use. This is great, love that. Thanks!
Sidenote: I think Weight 50 works better than 100 in most if not all cases, maybe iâm wrong
4 Likes