Blender 2.83 LTS | Hacker News Reader

Blender 2.83 LTS

blender.org

264 pointsdoener134 comments

I started using Blender around 2013 when we needed to make some exciting animations for a work project. I quickly grew to love it for its streamlined Jack of all trades approach to media creation and it’s breathtaking pace of development.

It’s such an incredible piece of software and I would encourage anyone interested in working with 3D graphics and animation to jump in. There is a wealth of training material online.

When you compare it’s basic workflows to other professional creation software such a as CAD, Photoshop the comparative bloat and lack of interface fluidity in the paid software is often jarring. It’s like comparing Vim to MS Word.

For example in SolidWorks if you want to rotate the view to a top, front or side view you can mouse over to a widget up in the corner of the screen or more quickly press space and then click a button there to orient the view. But that window has to be closed with the mouse and not by hitting space again and oddly can be resized for no good reason resulting in a big empty mess on so many users installs. It’s sloppy and janky design in extremely expensive professional software.

In Blender I just have to press 1, 3 or 7 on my numpad for front, side or top view as those keys correspond to X, Y and Z axes. Wonderfully intuitive and responsive.

In Solidworks the equivalent is ctrl-1 through ctrl-7. Random zoom outs can be caused by a misplaced part or sketch entity that’s way outside of what the user intended, which is very easy to do by accident. However I agree it’s a janky mess. Using the advanced features is indeed like resizing an image in word.

> In Blender I just have to press 1, 3 or 7 on my numpad for front, side or top view as those keys correspond to X, Y and Z axes. Wonderfully intuitive and responsive.

If you have a numeric keypad. None of my laptops have one; neither does my main machine. None of them have a middle mouse button either. The lack of those two things make an incredible amount of the Blender UX inaccessible to me. It's like making your site only work with Javascript and forgetting that some people have it turned off.

Which is why there is an option called emulate numpad under Blender Preferences > Input. The same is true for the middle mouse button (Emulate 3 button mouse). They both work fine, the numpad one is one of the first things I activate.

Not only is this option at the place where you would expect it to be, it is also not hard to find hints to it online.

There are checkboxes in the settings menu to substitute the number row and alt keys respectively

You may already know this, but there's an option in the Preferences menu called "Emulate Numpad"[1] that will make the numerical keys above the top row of letters behave as if they were on the numpad.

There's also a setting called "Emulate 3-Button Mouse"[2] that will allow you to use Blender with a trackpad or other setup with no middle mouse button.

References:

[1] https://docs.blender.org/manual/en/latest/editors/preference...

[2] https://docs.blender.org/manual/en/latest/editors/preference...

I’ve been learning it recently on my laptop. Use the Emulate Numpad option in the settings. The other thing I’ve done is remapped ; to numpad-/ (I think, whatever it is that allows me to focus on an object).

I suspect that ultimately you’ll just want a full keyboard because there are a bunch of variations that are probably useful to power users.

The other thing is that my MacBook doesn’t really want to render anything without also burning the house down, so I’m looking to build a desktop machine anyway.

Missing a middle mouse button is really easy to fix by using a €15 external mouse.

You can also buy external numpads that connect over USB. I have one of those.

Solidworks: Try Ctrl-1 through Ctrl-8

You can also modify the standard views, add views and select between Isometric, Dimetric and Trimetric projection.

The keyboard shortcuts are shown in the interface, which means it's easy to learn them over time.

I appreciate your enthusiasm for Blender. I also like it. My kids are learning it now through Udemy. However, I disagree with regards to the comparison to other tools. All of these tools have evolved over time to serve specific needs and audiences.

I've been using various engineering 3D CAD tools over the last 30+ years. There are some that can be painful to use and others which are easy and boost creativity. I would definitely place tools like Solidworks and Fusion360 in the latter category. Most of the tools that are painful to use are either open source or lack the financial backing (and profit motive/paying customer pressure) to evolve, refine and add useful features.

That said, the collection of great FOSS 3D tools out there is amazing.

Would Blender be a good tool to do mechanical engineering (for manufacturing prototyping)? Or would a tool like FreeCAD/SolidWorks still be much preferable?

EDIT: found an answer here, [1], but that was from 4 years ago.

[1] https://blender.stackexchange.com/questions/53293/is-blender...

If you're looking for a F/LOSS tool for mechanical CAD other than FreeCAD, try solvespace [1]. It's definitely good enough for simple usecases (especially 3d printing and lasercutting where there's good standalone CAM flows).

[1] - http://solvespace.com/index.pl

That looks nice. The only problem for me is that it is GUI-only (it does not expose e.g. a Python API, and apparently they are not thinking of supporting that in the near future [0])

[0] https://github.com/solvespace/solvespace/issues/375

Cool, how does it compare to OpenSCAD?

http://www.openscad.org/

EDIT: found an answer here: [1]

[1] https://news.ycombinator.com/item?id=9248701

Apples and oranges. Solvespace is closer to a traditional CAD package, with a layered constraint system, while OpenSCAD is more more mathsy and programming-oriented.

For purely mechanical parts where I don't have to repeat myself a lot, I found solvespace to be a much better solution.

looks like it was abandoned ...

No recent releases, but there's activity on GitHub right up to yesterday: https://github.com/solvespace/solvespace

That answer is still correct, which is that Blender's workflow and concept of 3D geometry is good for making nice pictures that don't need "correctness", but not what you want for mechanical engineering.

FreeCAD tries to be like SolidWorks/AutoDesk Inventor, and I think you'll find the workflow much better suited for designing mechanical parts in a precise way. Last I tried FreeCAD, it felt pretty alpha quality, but that was a long time ago.

Not trying to disagree with your point, but the company I work for built a design tool on top of blender that feeds custom designs to our automated mill. The customer gets a high quality 3d render to approve before materials are cut & packaged. It seems to work well. Maybe we are an edge case.

Are the items being cut all in planar sheets rather than organic surfaces?

eg things like a box, or IKEA furniture, assembled from flat sheets after being cut out.

Am kind of thinking Blender's lack of "proper" geometry might not matter much for that scenario.

Not really sure how it could be worked around well for more complex geometry though. :)

Yes sir

Blender is not a professional CAD software, but "good for making nice pictures" is understating its abilities; as a hobbyist I've modeled a dozen or so objects for 3D printing and they slice/print just fine.

There are even plugins to analyze models in Blender, and plugins for CURA to open blender files.

https://daler.github.io/blender-for-3d-printing/printing/mes...

https://thopiekar.eu/cura/cad/blender/

I'd suggest Fusion360 for this. Solvespace or FreeCAD if you must use OSS software. It depends a lot how complex the stuff you plan to build is.

Fusion360 is free for personal use so you can wait until you actually earn money with whatever you create until you buy a license.

Apparently, many users report that Fusion360 crashes often on Mac.

too bad it is Windows and Mac only

> Fusion360 is free for personal use ...

They've recently announced upcoming changes to that. Much of the Manufacturing (aka CAM) features are being put behind a paywall (US$60/month?), including some of the low-to-intermediate level ones like 4th axis CAM. :(

That's sad, do you have a link to more details?

I think this is the correct link:

https://help.autodesk.com/view/fusion360/ENU/?guid=Fusion_Ex...

The change was mentioned to me by one of our local Makerspace members a few days ago, who was considering learning CNC and asking about machines.

There's further discussion of the specifics (eg 4th axis) in the Fusion 360 forum somewhere. Found it a few days ago, but am not seeing it now. :(

Hmm, unlike "5 axis", it doesn't say that "4 axis" is part of the extension, it specifically says "Rotary", which seems to be this: https://help.autodesk.com/view/fusion360/ENU/?guid=MFG-MULTI...

On the "4th axis wrap toolpath" page, it doesn't have the same "this is part of an extension" warning: https://help.autodesk.com/view/fusion360/ENU/?guid=GUID-7FF2...

So I think (hope) that some forms of 4-axis machining are still free.

There's a forum post about it, where the Autodesk people say the 4th axis stuff is moved to the new extension. :(

But if you're actually machining stuff you're going to pay way more than 60/month in associated costs anyway?

Hell no. My local Makerspace has various low end CNC machines ranging from an old 800x800mm XCarve (3 axis only), that's useful for teaching people and using cut-offs from the wood workers.

We have a donated 4-axis machine as well, we've been getting up and running. Replaced the proprietary controller board with an Open Source one, and mostly just need to calibrate it and figure out an effective post processor.

If someone is letting you use their CNC machine for free, then sure. But whoever is paying the costs associated with maintaining that machine and buying materials is going to be paying more than 60/month.

Again no. I'm the person maintaining those machines. The maintenance cost is practically zero. Occasionally (every few months) we may need to buy a set of 1/8" bits (for $10).

Entry level CNC machines can be assembled pretty cheaply. eg PrintNC:

https://github.com/threedesigns/printNC

For materials, we have several wood workers in the Makerspace and can either use their off cuts, or buy a sheet of MDF ($10) that lasts a few weeks.

Doesn't it cost about $1000 to build a printNC? And if you are just cutting sheets of wood or plastic, you don't really need a proper CAD package.

The cost depends on the components you choose. Picking cheap components, and being careful, you can probably do one for ~US$500 (rough guess).

That machine would last several years. Though hobbyists do tend to upgrade their machines over time (eg add a 4th axis).

> And if you are just cutting sheets of wood or plastic, you don't really need a proper CAD package.

You might be surprised what you can create from MDF sheet, and/or 2-by-4" offcuts. :)

I have to agree with that. A decent CNC is a few 10k$, plus ongoing cost like tooling. At that point 60/month is pretty minimal.

Solidworks is pretty much industry standard, for good reason.

Blender isn't really appropriate for engineering. It's destructive editing and all a bit...wishy washy. Solidworks, by contrast, maintains a complete feature tree so you can see (and manipulate) every step that produces the final result.

My OP was not intended to suggest that Blender and SolidWorks were somehow competitors, they fullfill very different roles, just like Word and Vim, the point was about the janky nature of SolidWords and also a lot of ribbon infested CAD packages.

The cynic in me would think they deliberately obfuscate their software to sell training courses when in reality it’s more likely that development resources are constrained and the audience for the tools is small.

I think it's just the typical feature-based, got to have a reason to sell customers the new version, development.

Every new version (multiple thousands upgrade cost) brings a bunch of new features I don't use and make the software bloated. Yet, some bugs that annoy me daily are still not fixed multiple years later. That's what makes SW feel janky to me.

From what I heard, the original makers of SW left Dassault and founded OnShape which, despite being browser based, feels faster and less bloated. Or at least, so I've heard from people who've tried it out.

Hardly. They do a billion dollars a year in revenue.

I don't really think there is an "industry standard" for this kind of work and solid works is far from the majority.

You basic choices today tend to be Solidworks, Fusion360, Autodesk Inventor and in certain industries, Siemens NX. I have experience with all of these, as well as, in the old days, contorting AutoCAD to do 3D and 2D mechanical work.

I am going to say I run most of them at expert level, perhaps more so in the case of Solidworks and NX. I would not touch NX unless there was no other choice and I was required to use it. This is the case with a bunch of aerospace and other industries where there might be legitimate cases for using NX.

In the choice between Fusion360 and Solidworks, I would likely consider migrating to Fusion360 if they had a way to make it just as local as Solidworks. I haven't had to touch it in about two years, so I don't know if this transition took place at all. The fact that your data is cloud based doesn't sit well with many of the companies I tend to work with. In those cases Solidworks tends to become the go-to option.

From my perspective, yes, Solidworks is a standard tool. It is very rare --again, in my world-- to have someone ask for anything other than Solidworks when I ask what format they would prefer. This includes the new breed of private players surrounding the NASA Artemis mission.

I’m in that private player space and confirm solidworks is the most common, but I’ve run into the ones you’ve mentioned plus Solidedge, Creo, Ansys spaceclaim and a few others. We tend to use a neutral format for exchange. I’ve used pretty much all of them and I’d also pick Solidworks, but steer clear of anything beyond The standard package as the value is terrible and mostly doesn’t work well outside of a single person environment anyway.

Incidentally I recently did a study of what I should use for a side project (I.e. need to buy with my own money).

I decided that I’ll try Fusion360 (been a few years since I’ve tried it) and if it’s not enough I’ll try Solidedge. Solidworks doesn’t work For my use case (large up front cost plus annual fee). To restart maintenance/upgrades, you have to back pay for any years you missed, up to a certain point where you just have to pay the upfront cost again.

On the other hand you can pay by the month as needed (or not) for Fusion360 and Solidedge.

You might qualify for this; https://www.solidworks.com/solution/organization-type/entrep...

Not available in my country, sadly, and I don’t think I’d qualify due to the consultant/service provider exclusion.

For $500 US per year the Fusion 360 suite is unbeatable in my opinion. I have paid Solidworks tens of thousands of dollars over the years in annual maintenance fees. The simulation and CAM packages alone, if I remember correctly, were in the order of $9K US for the initial purchase and several thousand per year for maintenance for both.

What tends to happen with this kind of software --at least in my mind, not sure about how others thing about this stuff-- is that the ROI curve becomes asymptotic over time. The "support" fees don't match the marginal gains from release to release.

For example, I don't think we've had a need to contact GoEngineer (our reseller) for well over a decade. Once you achieve a certain level of competency there's very little they can contribute to the equation. Which means all that's left in terms of value for the "support" fee are the incremental software updates from year to year. And, as I said, with time, it's hard to point at features or solutions that justify the annual cost.

From my perspective if companies like SW/Dassault were smart they would progressively reduce the cost of annual updates to match customer realities.

The ONLY reason to upgrade past a certain point becomes interaction with other companies, vendors or consultants who might be on a newer version. Aside from that, I can't think of a single feature introduced by SW in the last many years that would justify spending thousands of dollars per year to get those features.

I haven't done the accounting at this point. If I had to guess I would say we have about $75K US per seat into SW and related tools. That's just crazy.

Don't get me wrong, I love SW. And, yes, we've made a ton of money using the tool. Yet that doesn't mean I am not going to be critical of a cost structure that, in my opinion, does not match the marginal improvements offered in exchange. It's negative value.

Here's a specific example from CAMWorks, the CAM tool we own and they push (to the point of SW having added a "lite" version of it for free with SW in the last few years). We have the full package, which wasn't cheap, along with training and maintenance over the years. And yet, for the tens of thousands of dollars we have spent on it, their tool, feeds and speeds library absolutely sucks (it will actually damage your machine if you are not paying attention) and they don't implement HSM (High Speed Machining) a critical productivity tool in modern machining. In fact, if you want HSM you have to spend (don't remember exactly) about $1,800 US for the initial license and (again, don't remember exactly) >$1K US per year for support and updates.

I can't wait to drop CAMWorks like a hot potato, it's horrible software we got stuck with over a decade ago. When you are running a CNC shop it isn't always easy to jump from software to software. People need to learn the tool and become productive with it. In addition to this, mistakes can be very, very costly (a spindle Z crash on a Haas VMC can cost you $5K to $10K US easy). In other words, you have to be very careful about the decisions you make and these decision become multi-year commitments, like it or not.

So, yeah, if Fusion360 had a full local mode of operation I would be throwing money at them. It's the right approach in my opinion: A reasonable annual fee, for a full-featured suite and modern software that seems to be evolving quite nicely.

I love SW, but it makes sense for you use case I would definitely go with F360.

> I would not touch NX unless there was no other choice and I was required to use it.

Is that from it having a poor UI, or is it more from a lousy attitude by their business/sales/etc side?

I think it's clunky and outdated in many ways. There are some things that drive me up a wall, like the issues with the selection of the units you want to use when designing a part or creating assemblies.

For example, in Solidworks you can change between units (millimeters to inches, for example) any time you want and as often as you want as you are modeling a part. In NX you have to run a script to convert the part file from one unit system to the other. It's an absolute pain in the ass.

https://community.sw.siemens.com/s/question/0D54O000061xYh1S...

https://www.eng-tips.com/viewthread.cfm?qid=235189

I haven't used NX in a about five years, I seem to remember having issues with mixed unit assemblies as well.

http://www.ata-plmsoftware.com/wp-content/uploads/2017/07/NX...

That said, if you are designing a rocket with 1,000 of your closest friends it seems NX is a better tool for the job. I've never tried to use SW in large team settings.

Bit of both... Solidworks is not cheap, and NX is about 4x of that. We're talking $5k+/yr/seat.

CATIA is pretty much used exclusively in automotive and aerospace.

Creo also has decent market share, though has lost a lot to SW and Fusion in the past few years.

No, they doubled down on the whole Fusion cloud thing.

Yeah, sadly it's a nonstarter for me for anything other than hobby projects. Even at that, since I own Solidworks there's very little incentive to use F360, even though I have it installed and really want to send them money (Solidworks and CAMWorks annual maintenance costs are ridiculous but I have no options).

For reference, this is where I directly asked about F360 supporting local-only workflow:

https://forums.autodesk.com/t5/fusion-360-ideastation-archiv...

Haven't really used it much since then, mostly because of their answer.

There are lots of tentacles to this cloud-based business. Autodesk is a very large and solid corporation, yet there are no guarantees that they are going to be around. I still valuabe AutoCAD files dating back 30+ years that I can open using a fully licensed old version of AutoCAD. The files are local, fully backed-up for decades, secure, private and we own the software licence. We don't need Autodesk to exist to access these files and use the software.

Same case with Solidworks. They could go out of business tomorrow and our current license would serve us well for years. And the designs would be safe and locally accessible within one computer, don't even need a network.

With clients, particular those in sensitive domains, cloud-based CAD is a nonstarter. I'm sure there are those who use it, I just haven't run into any yet.

Another example of this is the Adobe Creative Suite. We own full licenses from amny years ago. The minute they went to cloud-only we were done sending Adobe money every year. With permanent local licenses you can manage the financial decision of when and why you upgrade your software. With these cloud-based options some cut you off if you don't pay your annual fees.

I understand why these companies do this. I get it. However, they seem to be missing the point: Their customers use their software to create products and intellectual property. They don't use their software to use software. They could not care less. The value they create is in the work product, not the particular software they use. A model that cuts off access to your intellectual property (because you can't use the software) without an annual fee can flush years of work, decades, down the toilet. That's a terrible outcome. Again, there are no guarantees that any of these companies will be around forever.

JetBrains seem to have a reasonable model and we are happy to pay them on an annual basis. The massive difference between coding tools and CAD tools is that the work product isn't in a proprietary format that requires a license to that software. If we stop using JetBrains tools our C or Python code doesn't evaporate.

Yep, very much agree.

As a data point, I've investigated FreeCAD a few months ago. It seems "ok" for designing single part objects, and has reasonable 3-axis CAM.

FreeCAD's main drawback (from my PoV) is the lack of coherent assembly functionality. They have several incompatible, competing approaches all in development.

Hopefully one of those becomes the "blessed" approach, and people can safely use that for the next [decades]. :)

Did you evaluate A2Plus? Right now I'm building all my parts as independent Bodies in the same workspace, and haven't adopted A2Plus (since I like to design my parts within the assembly view, while it forces you to design parts then Import them into another workspace).

The purpose I was looking at FreeCAD for, was to publish some reference designs that would be useful both personally, and potentially externally for many years from now.

I tried the 3 current assembly workbenches:

* https://github.com/kbwbe/A2plus

* https://github.com/Zolko-123/FreeCAD_Assembly4

* https://github.com/realthunder/FreeCAD_assembly3

As they're all incompatible with each other, and none of them is the official "blessed" choice that's guaranteed to be part of FreeCAD for the next [decades], I just dropped FreeCAD as not yet being suitable for purpose. :(

Hopefully these separate efforts converge into a single system at some point, which would let FreeCAD compete with the commercial systems. :)

The rest of FreeCAD seems to be good enough (from my PoV), apart from that one critical blocker.

Any thoughts on Rhinoceros?

I've never used it. Don't know.

Frankly, if the cloud-based issue isn't a detractor I would not think of anything other than Fusion 360. It's a great product. Even the paid tier is amazing value for the money. Their $500 a year license would likely cost me well in excess of $5,000 a year in Solidworks. I mean, CAD, CAM and simulation alone get to that level in annual fees depending on what you have. So, it's 10x cheaper and likely a few x better value in some fronts. If they had a fully local option (with no cloud component whatsoever) I would likely switch away from SW over a few years.

It's a couple of years out of date, but see this:

SolidWorks has >2x the share of any other player.

https://www.quora.com/What-is-the-market-share-of-each-major...

Back when I was doing CAD work professionaly it was by far the most encountered software. Others have footholds in some industrys (aerospace, automotive) but for "general" product design SW is king.

That Quora post is a wrapper around CNCCookBook's 2015 survey [0], which is quite old.

They've done followup surveys [1], the most recent of which was in 2018 [2].

The latest survey shows that Fusion360 is now the most popular package and grew quite rapidly since 2015.

FWIW, Fusion360 is also the most popular CAM package [3].

[0]: https://www.cnccookbook.com/results-2015-cad-survey/

[1]: https://www.cnccookbook.com/cnc-manufacturing-trends-surveys...

[2]: https://www.cnccookbook.com/cnccookbook-2018-cad-survey-resu...

[3]: https://www.cnccookbook.com/cnccookbook-2020-cam-survey-what...

> SolidWorks has >2x the share of any other player.

That's a disingenuous way to interpret that data. Just 23% market share is pretty far from being an industry standard.

How is it disingenuous? 22.6 / 9.2 = 2.45

Onshape is nice, web based, built by the creators of Solidworks. Free for hobbyists.

Probably you want Fusion 360. Free for education or personal use, community driven updates. Think long term Autodesk wants it to be the future and the jank is all gone.

If anyone here remember's Star Fox, there's a fan-made series called A Fox In Space

https://www.youtube.com/watch?v=uieM18rZdHY

The story is, they moved all their assets over to Blender: https://www.youtube.com/watch?v=m09ZOHmGf8s

That's the whole kit. Apparently it's awesome at 2D-like animations. They started moving over before 2.80 dropped grease pencil: https://wiki.blender.org/wiki/Reference/Release_Notes/2.80/G...

The creator was updating his Patron saying the lack of updates was him spending a month mastering Blender. Now they don't have to hop around N number of different Adobe applications, with that cloud service running the bg slowing everything down. It even hangs when I restart.

Blender is great teamwork, open source at it's finest. But some bits of it - maybe just my imagination - feel it's a sort of wake up call to players like Adobe that overreach trying to push services cloud stuff on users.

By dropped you mean like delivered / dropped off, right?

Interesting that it can mean one thing or literally the opposite, with the exact same phrasing. Language is weird.

This is actually a common phenomenon, typically connected with meaning changing over time. Two examples that spring immediately to my mind are:

• Cleave: to cleave to something means to stick tight (this is now an uncommon usage), while to cleave something is to cut it in sunder.

• Let: now means “permit” but used to mean “hinder”. The older sense is broadly obsolete, but is still used in tennis (“let” meaning there was some hindrance to the serve, typically but not exclusively from the ball striking the net).

I suspect “couldn’t care less” being corrupted to “could care less” (which should mean something quite dissimilar) is a related sort of case, but I’m no linguist. I can imagine that all of these cases could be at least partly due to carelessness or misunderstanding of the word or idiom; but again, I’m no linguist.

Yep, added in 2.80!

Nice. Users have been screaming about how much changed between 2.79 and 2.81. They should have called it Blender 3.

The Blender 2.8x experience has too much of "where did they put ???". Remember, this is a 3D animation program capable of film quality, with UI complexity far beyond most applications. I spent 15 minutes yesterday looking for the "align" tool, which isn't where the "current" documentation says it is. That's a daily experience coming from 2.79.

Actually, the 2.8x series is pretty good. But it obsoleted most of the published Blender books, of which there are many, and most of the user base. So a stable version, to give users some relief and allow the documentation to catch up, is a big win.

I understand old users and experts may be annoyed at the fact things were moved, breaking their muscle memory (though I think this is at least partially mitigated by "classic" shortcuts map?). However please note, that for people who were always afraid and intimidated by Blender before, the new interface is an absolute game-changer. I finally feel I have some fighting chance to maybe try and slowly learn at least some of its basics, whereas before I felt completely alienated and frustrated even in the smallest GUI elements. More than that, I finally felt I can recommed trying to learn Blender to my close artist friend, who's not very much into computers, and whom I'd never dare to try and recommend the old version. Even with the new version I saw that for them it feels fairly hard and tricky to use, but now also definitively kinda somewhat familiar and maybe usable (the jury is still out, but I felt super happy when they since asked me "hey, so what was the name of this program you showed me?"). In my opinion, this is a seminal example of how to push a typically "nerdy" open-source project's GUI into a user-friendly one, bordering even on intuitive. I am amazed somebody, or the whole community (I don't know the story), managed to pull this out so well. And before Blender did it, I seriously didn't think an open-source project could manage to pull it out. If you take this into perspective, I think the changes to expert users' workflow, however painful, may actually soon prove to be relatively forgettable. If I am mot mistaken, I think all the old functions are still there; so after a period of frustrations of relearning the menus, I believe you're be no less productive than before. Whereas droves of people will suddenly be able to discover Blender and feel welcome into its interface, and empowered by the possibilities it opens before them. I sincerely bow with respect and amazement before the authors of this change.

I never got the hang of Blender in my previous attempts, but I started again with 2.8 in 2019, and the interface made a world of difference. In the past 8 months, I've gone from literally zero knowledge of box modeling to being able to create assets for indie games. I haven't seen any issues with the manual. Actually, combined with the official YouTube tutorials, I've found it extremely easy to find the answers to my questions.

It's still frustrating how different it is from all other software.

I know blender people are used to it but there are UI conventions. They work on all other programs. they fail on blender for "reasons".

One simple example. Click to select, press the Delete key to delete. Works pretty much everywhere. But no, Blender wants you to press X

Another is the way you move and or create window panes. Pretty much all other apps you just drag the title/tab/corner and if you to close there's (a) a close [x] icon, (b) no icon but drag the tab out to a separate window and one will appear (c) right click the tab for a context menu. None of these work in blender so it's off to search for a tutorial on how to close a window.

Another is using a non-standard file dialog which breaks all kinds of conventions. Sure they wanted more options I guess (though OS ones offer a place to add options). Or they wanted one UI across platforms, except of course people on each platform use used to that platform's conventions...

etc...

Still it's amazing software

>One simple example. Click to select, press the Delete key to delete. Works pretty much everywhere. But no, Blender wants you to press X

Both of these examples work as you'd expect in Blender 2.8+. You may want to try out a new version of the software.

File dialogue is also much more standard since 2.82. delete has worked as expected for many many (many) (since at least 2010) years

Having tried to learn Blender back in the 2.7 days, and having learned it this week on 2.82, the improvement in UX is almost impossible to overstate.

My kids are learning Blender on Udemy. One of them got started before the other and ended-up with v2.81. The other ended-up starting later and installed 2.83. He couldn't find buttons and certain functions didn't work as per the course he was following. He ended-up installing 2.81 to be able to get on with it.

I didn't dig into it so I don't know what the issues were. I seem to remember that object scaling wasn't behaving as it used to. We had two laptops side-by-side with 2.81 vs. 2.83 and they didn't do the same things. Not sure if this is normal with Blender minor version changes. Like I said, going back to 2.81 dealt with the issue, so no big deal.

Blender's versioning is a bit confusing. Change in the second number is not a minor release, but a proper major. They acknowledged it and gonna adopt more conventional versioning scheme for the upcoming 3.x [1]

3d software is a complex thing, many things could be unexpected for inexperienced users. I would suggest asking a question on Blender Stack Exchange [2] instead of downgrading.

[1] https://code.blender.org/2020/02/release-planning-2020-2025/

[2] https://blender.stackexchange.com/

Would love to see a feature where we can search for a button/setting, and Blender would highlight it. Just like search in DataGrip (JetBrain) for Mac.

This actually exists (press F3, with default keybindings)

Blender continues to impress me. I spent a semester my senior year of high school doing an independent study in 2011, then was able to do a project with it my 1st year of college. To my knowledge, every function is available via the Python API.

In 2011/2012 I worked on a project for a blind Chemistry student. MacMolPlt [1] shows electron fields and the physical location of atoms during chemical reactions. This visualization was useless to someone who was blind.

A CompE and I built a system which would take an export of the atom locations, bond type, and vertex location for the electron cloud surface and convert it to a 3d printed model. [2]

We came up with the following:

1) Import the Chemical Markup Language (CML) file. This is XML and contains basically all the information needed for the screen visualization.

2) Each column of the periodic table was a certain shape.

3) Each row of the periodic table was a certain scale.

4) Each bond type had a different "rod" connecting them.

5) The electron cloud surface was imported.

6) Everything was "unioned" together to make a surface model.

7) The file was exported to an STL file. We then submitted this to a 3d printer.

I asked the professor I was working with about making it open source. The answer was basically "we might be able to do that, but it's going to be a political mess at the university". The money to pay me was from the disability office, with relatively little oversight. They wouldn't necessarily object to how the funds were used, but they would probably have an issue paying for enhancements. We weren't sure if they'd want to recoup the costs. The market of beneficiaries was very small.

[1] https://brettbode.github.io/wxmacmolplt/

[2] https://imgur.com/a/kxeDjy6

One thing I appreciate about the Blender Python API is that it follows the conventions of the wider Python ecosystem [1].

Other tools, like KiCad and FreeCAD for example, have Python APIs, but they tend to follow the conventions of the underlying C++ code. In the wider scheme of things it's pretty minor, but it the consistency is nice.

[1] https://docs.blender.org/api/blender_python_api_2_63_7/info_...

> MacMolPlt [..] shows electron fields and the physical location of atoms during chemical reactions

ASE and GPAW might be another option. Python, Linux/Mac/Win, with "optional" GPU accel.

I'm exploring teaching atoms and materials less wretchedly, down towards early primary. Using ASE/GPAW for trajectories and electron densities, for physically realistic interactive visualizations in XR. No complaints.

https://wiki.fysik.dtu.dk/ase/ https://wiki.fysik.dtu.dk/gpaw/

I love what blender has done with the interface.

It's like the exact opposite of Inkscape, which gets worse by every release. It no longer looks like incompetency, at this level I suspect sabotage by a competitor...

And GIMP.

I use GIMP and Inkscape with some regularily. I wonder if I should be using Blender.... it seems way better but also does so much more.

Does it replace much of GIMP/Inkscape functionality, or is it mostly just for 3d and animation?

I'll second Krita - it has amazing functionality for both vector and raster editing, and mixing the two in the same document. Haven't looked back since I switched from GIMP.

it can replace gimp but it's not made for it. I'm not sure about how much the new grease pencil adds for photo editing though.

Yeah well I don't just use GIMP for photos, but as a general-purpose 2d paint program. Any time I need a graphic on a web page or post or what have you.

Krita is worth looking at for this sort of use case

My son and I are learning Blender together. I was an avid 3DS Max user for many years (Kinetix, then Autodesk), and Blender is both impressive and a comfortable next step. Never did much with Maya, though there was a detour to Softimage for a few years...

With a GTX 1070 video card, Blender has been quite responsive on a dual-monitor HD setup. Highly recommend that users try the Nvidia Studio Drivers. We started off using the latest game-ready drivers and the system was unstable, but after doing a clean install of the studio drivers (v 442.92), its been rock solid no matter what we throw at it.

It's absolutely crazy to me how fast the sculpt mode has improved, especially since it's mostly done by one guy!

It's not going to "catch up" with zbrush, but it being integrated into a complete 3D suite is a huge boon. It's still missing some key features (e.g. polypainting, which is coming soon I think), but if it keeps improving at the current rate, I could see it being a serious contender especially in areas such as concepting.

For artists who can't afford ZBrush, I believe this is a gamechanger.

Blender needs a remeshing tool on a par with ZRemesher, however there is Exoside Quad Remesher which is by the same guy as ZRemesher and offers pretty similar results, although its over $100US. The new Quadrimesh in Blender 2.8 is great but it still doesn't quite hold a candle to ZRemesher.

Blender has come such a long way since it went open source. Love it, great work, and thanks to all contributors!

Ha! I didn’t even knew Blender was closed source at some point.

It was originally a commercial product. The seller went out of business, and the user community bought the software from them.

I remember contributing some of my pocket money at the time!

Me too. The success of blender really sold me on Open Source.

It makes me happy to see the speed at which Blender is being developed now. I have been following it for more than 10 years and it seems like it is finally starting to take over the industry. :)

And even though I don't use it that much myself, it is one of my favorite FOSS projects out there. I wish Blender developers and community the best!

I do architectural modeling as a side thing. My needs in more marketing materials rather than construction documents. Would blender be a possible replacement for Sketchup given the recent announcement that SU is going subscription only?

Absolutely, it would be a most suitable replacement. I've being using it for similar purposes, as a professional architect, exclusively since 2013 or so.

Blender delivers great results, is a joy to use and has a vibrant community.

Blender’s cycles renderer and some HDRIs would be an excellent way to produce stunning architecture rendering. Switch to the eevee renderer and use freestyle to outline edges (which you can also export to SVG!) and you can produce some excellent stylised images too.

Also have a look at the included Archimesh add-on.

https://docs.blender.org/manual/en/latest/addons/add_mesh/ar...

+ also Archipack (bundled as well)

One of the thing that has always amazed me with Blender is the sheer speed at which it saves and loads large and complex scenes to/from disk.

I believe none of the other 3D packages is anywhere near that.

I'd love to hear how they do it if anyone knows (or has pointer to an explanation).

I couldn't find very recent documentation, but the Blender wiki archive has a few articles that explain saving files. This article [1] in particular was an interesting read. It appears it is mostly a direct write of structs to disk.

[1] https://archive.blender.org/wiki/index.php/Dev:Source/Archit...

Somewhat related, check out Blender DNA. It works similar to Protobuf. I remember reading a pretty good article on it, can't find it now.

This release has a well needed update to the gLTF exporter which is great as that's the preferred method of importing assets into Godot.

It was possible to update the exporter plugin separately via Github before but this will help a lot of people.

Great news! gLTF is also a preferred way to export models for use in webgl 3d frameworks, like three.js

Also becoming that way for Unreal too, they have had an importer since 4.19

The native support for OpenVDB volumes is exciting. Having worked with OpenVDB occasionally, it's a very impressive library. The speed at which it can execute complex volumetric operations is matched only by the difficulty of understanding how to correctly invoke them. OpenVDB's points -> level set -> mesh is an order of magnitude faster than the marching cubes implementation I'd been using previously.

OpenVDB is a hidden gem, and I hope this gives it a bit more exposure.

I recently got into 3D printing and found I often need to modify models to get better prints. This has lead me into learning some modeling basics.

Started with fusion 360 (because it’s free), but even as a new user the entire interface is clunky and hard to use.

I then switched over to Blender 2.8. To be honest, it’s not a simple set of tools for a beginner. However there are tons of great articles and videos about how to do things. After a few weeks of working through simple tasks and then on to more complex things, I’ve grown to love Blender.

Once you learn some concepts and get a handle on workflows, I can honestly say Blender is a much better tool than fusion 360. Not only is it more powerful and faster, the folks working on Blender have clearly put a lot of thought into usability. If you’re frustrated with Blender as a new user, just keep at it and you’ll pick it up quickly.

To add another point, my experiences are specifically oriented towards modeling and modifications for 3D printing. You’ll still probably want to use tools like fusion if you’re doing other tasks like mechanical design.

I'm extremely surprised you found Fusion 360 hard to use. It's one one simplest CAD software I have used. I've used in the past Pro/E and Solidworks, and Fusion 360 is bar far the easiest.

I’ve used lots of CAD packages over the years and while F360 is pretty easy it’s also a pain with lots of tedious behaviour like having to upload an STL to import it.

Does Blender still export complex geometry with holes in it (eg STL files), that need manual fixing before they'll print?

As mentioned, enable the 3D printing add-on (it’s built in) it’ll check your mesh and highlight trouble areas.

https://youtu.be/-y5r2mTlUWA

Cool, that looks like a reasonable workaround. :)

Of you've modelled it so it has holes in it then yes... This is definitely one of those where it's on the artist to make sure the mesh is manifold, watertight and orientable because you want to turn it into a physical object. There is a nice voxelise feature and some 3d print tools that will make things easier and fix minor errors but they won't help you if you try to export something that is badly modelled.

Any advice for getting started with Blender? Been making 2D games for awhile and want to jump into practicing some 3D, of course that comes with learning to model.

I've had a great experience with two paid resources: CGCookie.com and blendermarket.com. CGCookie is a monthly subscription site that has courses on modeling, materials, sculpting, and a lot more. They just refreshed their course on mesh modeling for Blender 2.8x [1].

Blender Market can be hit or miss since it is user-created content that varies from really amateur to quite professional. If you want to learn the ins and outs of shaders in Blender, I highly recommend these two courses from the user "Blender Insight" [2][3]

References:

[1] https://cgcookie.com/course/fundamentals-of-3d-mesh-modeling...

[2] https://blendermarket.com/products/become-a-material-guru-in...

[3] https://blendermarket.com/products/all-about-eevee-for-blend...

I think Youtube tutorials are the way to go.

Blender Guru is the most popular channel, but I personally prefer the tutorial series[0] by Grant Abbitt for getting started.

0: https://www.youtube.com/watch?v=7MRonzqYJgw

Plus one on Grant, he does a lot of livestreams too and critiques viewer submissions.

I just went through Blender Guru's donut tutorial and it is excellent as it covers a whole progression from learning the basic UI to modeling, sculpting, texturing, shading and even a touch of animation.

For a change of pace, look at Ian Hubert's "lazy tutorials". Each one is only one minute long but simultaneously shows something amazing while making it seem like anybody could just pick up the tool and do it.

Thanks! Those Ian Hubert videos are amazing

YouTube tutorials. There are many of them. It's great to follow along at first since they'll teach you about the UI.

I did the CGBoost beginner series last summer. It's a paid course but because if that it's much more comprehensive and higher quality than a lot of the free stuff on YouTube.

It takes you from zero to making an animation of a sci fi car traveling through a desert.

For anyone who finds the new Grease Pencil tool exciting but intimidating, a YouTuber named Dedouze (whose amazing artwork is the Blender 2.82 splash screen) has a two-part course on drawing and animating with the grease pencil.

Part 1:

https://www.youtube.com/watch?v=c57qq2nE3B0&t=1404s

Part 2:

https://www.youtube.com/watch?v=mhcNPYxPqhw

The Blender Foundation team is absolutely killing it, not only with their product development, but also their marketing.

Which reminds me, I'm long overdue in donating to BF. So much inspiration and utility over the years.

Does anyone have experience using blender for ML simulation data? How does it compare to Unity? I’m using Unity for generating synthetic images and RL learning.

What kind of synthetic images are you wanting to generate? Blender has a lot of power in terms of procedural generation.

One example is chess positions. I render them in Unity and take screenshots to determine if my ML architecture and loss function can read the board from just and image.

Well I know you could definitely render the same thing in blender and it would probably be faster as blender is actually designed to render images and animations, as opposed to Unity being primarily a game engine.

As far as generating them procedurally I'm not sure. I'm assuming you're using 3d models of the chess pieces, and while it is possible to do procedural animations with the animation nodes add-on I haven't used it much myself so I'm not sure if it will me your use case.

https://github.com/JacquesLucke/animation_nodes

Blender also has a great python API. AFAIK it has an API for every single function so the sky is the limit really. I don't think it would be too much effort to create a blender file that on startup would execute a script that generates a random chess board arrangement, inserts the corresponding models, and sets up the camera.

Blender can even render from the command line, so you could probably just setup a loop that keeps rendering the same file over and over to generate a whole bunch of positions.

I see a big NVIDIA part on that page and wonder where in the world AMD is. Are they part of blender? Because maybe they should be.

Reply on news.ycombinator.com