Per the first bullet point in the OP, the workaround for this which I’ve added to our internal documentation where I work is the following rather silly dance:
Link the collection containing the hierarchy you want to work with
Perform select hierarchy on the top level object in the outliner
Hit the shortcut for move to other collection (which will link it instead)
Unlink the linked collection
Work on the linked hierarchy in your new local collection
If anyone has a better suggestion I’d be interested to hear about it!
Also, if anyone knows of an add-on which can list all linked objects, show where they are linked from and individually re-link them, I’d absolutely love to donate/buy that!
It is true that the BF has never done a sterling job at polishing up major new features in subsequent release cycles (with some not seeing much in the way of polish and bugfixing at all).
Fixing this issue though is easier said than done, because in part it relies on fixing another major issue with development and that is patch review and doing what they can to help volunteers get their work in. That and the realization that you are trying to resolve a habit that has been embedded in Blender development since version 2.25 (nearly 20 years ago).
Alas though, the only thing that would prove to be a potent alternative to Blender is a fork of Blender, and that is simply not going to happen without having a well-funded team right out of the gate.
There’s a GSOC currently underway that’s refactoring the outliner; perhaps somehow this can someday result in improving instancing, linking, and grouping in ways that are desperately needed.
Well, I just threw in the towel on our current project.
I could sort of live with material overrides resetting, and the finnicky UX, but I encountered a hard crash which I can’t get around (wasted several hours trying).
Everything is now local and I declare library overrides to be not ready for production and will warn my coworkers against using them.
Oh, and since the developers don’t care about UX, allow me to vent a little bit here:
While making everything local, I had to break some links to avoid the crash. However, empties with broken links (how that even is a thing) can’t be bulk selected in the outliner (using filters) to make use of “make local” context menu (or the one in the object menu). You have to select each. Individual. Empty. And. Click. The. Link. Icon. And. I. Have. Hundreds.
Maybe if you have dedicated developers like this company you can work with it, but if you’re a small team of just a few people, forget it (actually, at 9:22 he mentions that he can’t override it either)!
I hate the link and override problems, but heartened to see I’m not alone. I thought I was the only one.
Thanks for the info on Asset Browser’s thumnails. Yeah. Not at all intuitive, and needlessly complicated too. A ‘screenshot current view as thumbnail’ button would have made more sense.
Thanks for sharing that, @eobet. Unfortunately that only works to one level of the hierachy, so you can’t get the stuff buried deeper. I’ve “solved” this by either (a) making my collections so anything I might want to tweak is pushed up to the first level of the collection hierachy, (b) hiding first level objects and then creating link to the now hidden object directly in the library, and repeating the process as deep as is needed. That’s difficult, because it gets back to the problem of not being sure what you’re looking at on screen, as well as putting it at the right location. The ‘Edit Linked Library’ Addon can tell what the linked collection is - though unfortunately not the object. I wrote a script that identifies object clicked on, though it still requires a lot of trial and error. Half-an-hour is pretty typical to pin something down this way. Something that should be doable in seconds.
Considered writing a script to overcome this, though it’d be a band-aid on an ever changing tumor.
I think it’s like this:
If an established animation studio with a say Maya pipeline approached me saying they’re considering switching to Blender, I’d recommend they don’t, not just for the reasons in this thread (which are substantial), but also loss of expertise and risk.
If a start-up animation studio asked me if they should go Maya or Blender, I’d recommend Maya, because you have a far better chance of completing your project without running out of money. In Blender I’m encountering problem after problem and that really hurts productivity, which makes Blender far riskier proposition. You might lose your shirt.
Who does that leave? Hobbyists, and start-up animation studios who use Blender despite the risk because of the cost. Neither flush with cash.
There’s also a matter of what your donation goes to. BF seems to opt for “Cool” additions like Geometry Nodes and Grease Pencil. I do understand some people are really into Grease Pencil, but personally, I think that development resource would be better spent elsewhere, especially since there is already the free professional-grade open-source OpenToonz animation system available (Futurama and Ghibli). I do appreciate Geometry Nodes, but not at the cost of ignoring fundamental core function.
The bugs and crashes are a real problem too. I’ve seen is-isn’t-is-isn’t Intenet arguments on whether Blender is buggy. Think it depends on what you’re using it for. On a high-end PC, I find Blender is largely bug free for polygon modeling and sculpting, but on a low-end PC (integrated Intel chipset) it crashes frequently. For animation, even on a high-end PC, it crashes frequently. I kid you not: I get 10 to 30 crashes a day. They’re annoying, and they really interrupt the workflow. They are also unreportable, because they’re usually not repeatable. If you reload from your auto-save and repeat it, usually it works. As a rough comparison, Maya crashes, but maybe once every three days.
The shortcomings we’re describing here, and documentation, testing, bugfixing: These are boring by comparison to new and cool, but IMO far more important.
Of all the software I use, Blender is by far the buggiest. The bugs and crashes seen are indicative of lack of testing. Low-end PCs, because BF doesn’t use them. (When someone reported bugs in the integrated Intel chipset, a Blender Dev replied they hadn’t tested Blender on that - despite its popularity - because they don’t use them), and animation on High-end PCs because only a small number of users do animation, and animation is very likely to flush out bugs because by it’s nature you’re generating a huge volume of work and effectively stress-testing the system in a way small projects don’t.
For corporations, bugs and crashes are the kiss of death. Any company which trials Blender and sees that will walk away without a further word.
Worth pointing out a difference here: For animation, everything has to work. For game asset creation, you only need enough to work to be able to export your asset, then you’re done. IME for game asset creation, Blender is great. For animation production, no.
Blender Studios has a different view: “Overall the feedback – mainly from the Blender Studio artists team – over the current library overrides was very positive. However, several critical pain points and potential improvements were identified, such as performance during resync.” https://code.blender.org/2022/02/overrides-workshop/
Me, I hate overrides in their current form. If you’re doing simple rig animation, yeah, they work, but if you’re doing anything more sophisticated than that, they just don’t cut it.
Yes. Talked about that earlier. I found reporting bugs to BF to be a waste of time.
Yeah. Apart from the linking crashes - which often aren’t repeatable, the UX problems discussed here aren’t considered bugs. You’ll often hear ‘Post it in RightClickSelect’, but there are good ideas which have been sitting there for years without comment or action from the people who matter.
Been there, @eobet. The fix for dysfunctional overrides is often to make it local anyway. Have you Binged for python code to do that on the outliner, or just select the objects and run a script to do it with ops?
That’s the big question.
I don’t think forks are feasible. Blender For Artists has a tiny share (0.8%?), and you need to be able to pay your developers, but where is that money going to come from? I can’t see it. An alternative is to persuade BF to shift their priorities. Can that be done? Also a question as to whether they think it matters? Back in 2013 Ton said he didn’t care about attracting Maya users, since Blender already had hundreds of thousands of users. Would BF be willing to alter course to attract animators and future animation studios, even though it’s a smaller niche?
I’m committed to finishing this current project with Blender, but if BF isn’t interested in improving its functionality, maybe using Blender for asset generation and Unreal for animation and rendering is a better idea?
It’s an interesting testimony you made here ! Since I have a very different experience with blender on animation projects, can you give more insight on what kind of projects you’re working on, the scale of the team and how blender fit into the pipeline so it’s easier for me to understand where the difference come from ?
Many studios / projects I’ve seen using blender (generally small to medium sized one) on animation projects seemed to be fine with it, very far from what you are describing.
However, for blender to be effective, studios need TDs specialized in blender with a good knowledge on how it’s supposed to work. Which is hard to find.
It occurred a few time that I saw CG supervisors trying to recreate how they work with Maya within blender, and while they might get close to that, it ends up in a very convoluted organisation.
The closer you stay to how it’s supposed to be used, the better it works, despite some annoying limitations.
You can show some companies testimony here :
All that said, I’m not sure blender is ready for 100 million $ movie projects anyway, it always has been targeted at small/medium teams.
I can’t speak a lot about using overrides since I’m much more used to the old proxy system, however :
It’s logical, there isn’t a good alternative to that :
In green it’s the original data, in red an instance of “Collection”.
Instance is at (4,0,2) but what is the position of the original data ?
If I make an override of the collection in red, I indeed expect the red cube to stay in place, that would mean blender would have moved the content of the collection to (4,0,2). In that particular case it would work. But many advanced rigs would break if you move armature or helper objects out of their original /intended position. These rigs needs to be moved in pose mode rather than in object mode. Since override is used primarily on rigs it’s understandable that it works that way.
There isn’t a lot of alternatives : everything can be parented to an empty, therefore object keep their local transform, only the master empty is moved.
Solution that could work but worse than the issue would be to induce and invisible offset, like delta transforms, or to let advanced rigs breaks, all of which would lead to complains at some point anyway.
Best practice on big scale project is to use a custom asset loader that ensure each rig is loaded with their collection at 0,0,0 and the rig is overridden from there. That way layout artists put the character/props/set in place using their rig and they don’t do the override, which should be handled by pipeline scripts anyway.
It seems you’re mis-using the asset browser, which should be used for managing local assets only, it’s not meant to be used by several artists on the same project. A good use case is if you are a modeler and want to use you own library of kitbash pieces. It’s planned on the roadmap that Asset browser could handle “projects” but there is still stuff to be done before.
In the end, you have to build your own asset loader that will handle referencing, versioning and loading/overrides of assets according to your database ( shotgrid, kitsu …).
It’s not so much about attracting Maya users as it is about adding value for existing users, you simply don’t know if Maya users are going to switch. So, Blender develops for their users, like Maya does for their customer base (they don’t add Blender features to attract Blender users either, roughly speaking). If a new feature appeals to both existing and potential new users, that’s just added benefit.
Indeed, I think since blender is open source it doesn’t really depends on higher number of users but rather quality users that will help it grow.
Some examples of studios contributing to blender :
Problem is that it’s sometime a chicken and egg issue, studios are welcome to contribute to blender if something is missing, but studios might not be interested since blender isn’t 100% ready for their pipeline.
But, for those who use blender for a long time this can be seen as a good thing since blender is making progress anyway in the long run.
When I started my career blender was used only by individual who need a bit of 3D in their mograph/VFX work. Then it evolved and small team based 3D projects where more and more common.
And year after years more and more bigger sized studios are starting using it.
In France many schools start to teach it alongside Maya because more and more studio use blender in their pipeline.
Maybe at some point it would be the right tool for more advanced animation project, at least it feel sane to me that BF isn’t forcing into that and that it comes as a step by step process since it’s a huge task anyway.
Let us not spread myths about what the BF’s reasonings behind the development of certain areas are.
I’ve heard this same argument before (likely mainly from sculptors who feel left behind in terms of development).
With that said, I don’t think it is fair let alone accurate to pull the hypothesis there’d be some agenda behind this out of thin air. Something along the lines of 'geometry nodes are cool and flashy and showcase-able and twitter-able and marketing-friendly, so let us put large efforts into that for the sake of PR-benefits'.
I’m pretty sure such reasoning is not the source of BF’s constant work on geometry nodes.
Instead I think a major factor is simply the fact geonodes are an alltogether new system which is being pretty much designed from the ground up as we speak.
So there is very little legacy code to be factored in, very little need to worry about not breaking things already in place, working within limitations of systems already established etc.
I believe it should be obvious such work is much more straightforward than if you’re facing lots of code already in place, may have to do refactoring and so on, like is probably the case w.r.t. something like the rigging and animation module (or the sculpt and paint module, for that matter).
After all, you said yourself:
Also remember, much work on geonodes was done by volunteers, and so was work on greasepencil (especially in the beginning, iirc). Look at the list of attendees from the last GreasePencil module meeting
Summary
For comparison, the list of devs currently on the BF’s payroll (source):
Summary
That leaves Falk David as the only name common to both lists.
.
.
Do not get me wrong, I do agree with much of what you’ve written, @Conan_The_Blendarian. Just please don’t make up rumours about geonodes or greasepencil receiving attention because of ‘cool’ (whoever claimed that).
greetings, Kologe
edit: Put lengthy quotes into collapsible spoiler-thingies for brevity/readability of the post.
Tangent studio collapsed when it was forced to switch to Maya by Netflix in half-year.
They was using last DMC4 massive collaboration compatible stable version - 2.7x.
Since then maya users requests are heavily prioritized, and software quality has gone straight to DMC1 personal use level (a famous 2.8 design quality dropdown), since features design are prioritized over system design, resulting in software development style similar to Autodesk’s which always has been mostly about marketing and satisfying users feature run.
It is in general hard to remember anything finished properly in Blender’s system design after 2.8 quality dropdown. Previously it took years to find a critical system design error, now it takes minutes.
Our studio found handling Blender-based production extremely expensive at that point because of lots of infrastructural flaws and general unpredictability.
Well, sure blender as to adapt to it’s rapid adoption by many studios.
I’m not sure they blindly prioritize requests from Maya users, but they may try to fit better in the entertainment industry.
Well, it’s very possible that blender was more stable at some point, and currently all the massive structural changes surely don’t help.
But still a lot of things where kept WIP, IRRC BMesh was supposed to completely replace the old mesh system and finally they kept both, or the particle system which was heavily developed at some point and abandoned since then. I’m sure it’s possible to find other examples. Since at some point blender was mostly developed from contributions it’s logical that people come and go letting projects unfinished. But maybe since it was a smaller codebase it was easier to stabilize it anyway. TBH all that is a bit over my head since I’m not an engineer.
From my user experience POV, for the most part blender become better and I don’t feel it being way more unstable than it was. But I believe what you say since you probably have a very different experience with the software due to another way of using it/ different application.
I watched the Cube Creative Studio presentation. If they have used Maya/Max, they wouldn’t need to spend resources on developing LOD system. In the end, they saved a panny, lost a dollar.
Tangent also had a big pool od developers. The salary of just one dev would more than the lic cost for thr whole studio. What they have used is not same as, what you have.
Well, they use max before since a very long time, switching their whole pipeline to blender was a huge work ( took them a year to build their first pipeline for 2.79).
I’m pretty sure if that means loosing money in the long run for them they would go back to max. If they stick to blender that means that they got at least similar costs and they find more beneficial to stick with blender. It’s a big company they won’t stay with blender just because it’s cool.
Jeff Bell of Tangent was an Alias/Wavefront Engineer that worked on Maya (years before Autodesk bought it)… 4 months after shutting down Tangent, their in-house production management software was bought by Autodesk and several key team members hired.
It doesn’t sound like “Netflix made them use Maya” is the whole story…?
