I am no developer, but I have a question regarding simulation nodes. When are they going to be implemented? There are alpha builds, but I do not want to stuff up my works with unstable builds. I like to wait until things get more stable before trying, as it is not worth the risk.
They are in main for 3.6 alpha, so probably late June unless something drastically changes.
If you want to check them out you can always test out 3.6, but continue using your current stable version for your work. You arenāt limited to one build only on your system.
2 Likes
You can read the Monday meeting reports, theyāre very understandable. The team is working on caching and UI elements right now. First release will probably be pretty basic, theyāll keep adding nodes and options in further releases.
Iām curious to see when it can replace rigid body sims, which can be very unpredictable. But I also fear that sim nodes will be way more complex to master.
2 Likes
The initial release of simulation nodes was going to be in 3.5, but it got pushed back to 3.6. In theory it will make the final release of 3.6
Keep in mind, this first implementation is more a āmotion graphicsā type of simulation for geometry nodes. It wonāt work for Cloth or Hair in 3.6.
The earliest release that could have Cloth and Hair simulation nodes is Blender 4 at the end of the year. But even that is only a maybe at this stage.
1 Like
Nevertheless even with this version it is possible to do cloth-like simulations (although with a lot more effort); check Seanterelle at twitter.
For me one of most important things is when we can get nodes of nodes aka Compounds like in Softimage ICE.
1 Like
Yes, it is going to be confusing for many people in the early stages as they will expect solvers to be built in. 3.6 will only be the bare basics of the simulation and caching framework. You basically have to build your own solvers to actually do anything in 3.6 sim nodes, so users need to keep their expectations in check.
I have played around with some particle and very basic dynamic sims in 3.6, but am not interested in trying to create a proper dynamics solver. Most of what is being shown in the community are simple point based dynamic solvers., which they have built themselves.
I am guessing that 4.0 will have some form of particle nodes, as the aim is to get rid of the current particle system for 4.0. The good stuff wonāt arrive till 4.x when Lukas Tonne and other devs start hard coding some atomic solver nodes.
4 Likes
Iām really hanging out for full hair simulation and the hopes it is better then what we currently have.
All those new and included hair node groups are great, but I just fear how much it may half fall apart when it comes to animation and art direction/manual adjustments.
So far Iāve avoid looking at the simulation nodes at all, since I expect a lot to change as they try to sort it all out and till there is any real chance of doing hair sim, then I have plenty of other stuff to do.
Yeah, to be honest I am a bit apprehensive with that whole bundled node group approach. I totally understand the reasoning behind it, and Houdini has been going in a similar direction for years. But still, when your whole production is relying on someone elseās premade node group that could easily break from release to releaseā¦hmmm, it does make me somewhat nervous!
I guess time will tell how well this works and how well those node groups are maintained. I think you will see hair physics fairly early on in the 4.x cycle as this would be a high priority for the blender studio short films.
Other short term solvers will probably be for the particle system and gaseous fluids. I just wish they had hired that guy that was doing the amazing looking gaseous fluid solver on devtalk. He approached the institute for a job and was turned away. They could have at least got him on a short term contract or something.
Not really sure what the plan is with solvers, and if they are going to use an open source library or start from scratch. If it is the latter, then we have a looong wait aheadā¦
I do agree that a suite of higher-level nodes will be needed (because it wonāt matter to artistic-minded users if they are a bit more of a black box compared to node groups that can be modified). This assumes they will exist alongside the lower-level nodes by the way.
Fields are far easier to get your head around than the original design with all of the typing, but it can still be pretty complex stuff that takes a lot of time to use correctly. Sure, it becomes a lot easier once you build up an asset library, but Blender also needs to have provisions for those who just want to throw a scene together even at the cost of fine-grained control.
Gates of circus have opened 
, I am sure a lot of smart people will create a lot of good solvers, I was fooling
with Seanās demo file.
2 Likes
Node groups ?
Yes, did not know it was possible to put in asset browser. Still it is inconvenient to share with need to share blend files and reuse without using asset browser.
Not familiar with ICE unfortunatelyā¦ but as far as I was able to collect they are pretty much node groups ? Do you wish they were shareable independently of blend files ? as json or something ?
Seeing how well-made Essential Hair nodes were in 3.5 Iām now more optimistic about that workflow. If Blender takes some time to create those types of nodes for simulations and solvers and maintain them it could be very comfortable
2 Likes
Yes, you could save any group of ICE nodes as compound which if it was in a designed Softimage ICE directory was available in ICE tree menu but you could also share that compound since it is a normal file recognized by windows. So if i have an effect i want to share with another person i could just send it by email etc. When you save a compound you also save versioning of said compound can append notes etc.
Some user can just shere a compund like this
1 Like
Blend files with just nodes in them are pretty lightweight so you can share them just as well. I donāt know of any plans to allow exporting a nodetree but I recognize how that could be handy.
1 Like
Yea basically i can just share dozens of them by just selecting them in explorer and sending by email.
Another option when saving a compound is also the classification. So if it simulation it would appear in simulation folder of Ice Tree menu etc.
Yea I think asset libraries are supposed to handle this. It has been made clear that the design was thought for āone asset, one blend fileā, but I personally store several utility node groups in a single asset file I call utils. You can link them in too so theyāll dynamically change when you update them in the source file.
I canāt see that being a good path to promote nodes use that is already complex. Softimage ICE system is much better