[raincole]

It does feel like Blender is eating the world of 3D content creation.*

I know the industry standard of animation is still Maya, and geometry node isn't as powerful as Houdini, but Blender has made such great progress so far... I wonder if there any hobbyist/student who learns 3Dsmax or 3DCoat as their first DCC anymore?

*: Assuming genAI won't deliver and make that world disappear entirely, ofc.

[mrandish]

> Assuming genAI won't deliver and make that world disappear entirely, ofc.

I get the concern and the reasoning behind it but having spent most my adult life and career in digital content creative tooling (2D, 3D and video), I believe AI will be short-term disruptive but also a long-term net positive. It should make skilled pros faster and more productive while helping non-pros achieve more of their creative vision on their own.

For over 30 years progress in tools has been about giving creative people the power to realize their visions. To create more and do it faster, cheaper and at higher quality. Of course, they don't always choose to use that power well but the concern that "these new tools are just enabling bottom feeders to create more bad content" has remained a pretty constant refrain since I started in 1989. Even back in the 90s I said "Sure, these new tools will enable 95% more crap but they'll also unleash 5% that's great which wouldn't have existed before." I think that's just as true today as it's always been.

[mschuster91]

> It should make skilled pros faster and more productive while helping non-pros achieve more of their creative vision on their own.

The problem is, AI is good enough to replace juniors. That means companies are already cutting positions at that level and some are just itching to ditch intermediates as well once quality improves.

But when juniors and intermediates are all gone... how are the beancounters expecting to get new seniors when the old ones go to retirement or hand in their 2 week notice because they are fed up cleaning up after crap AI?

[mrandish]

Disruption is disruptive and can suck. But it's also not new. Some companies and people will do short-sighted things in response and pay the price. Other job types and roles will simply be disintermediated and those individuals will need to learn new skills and find different kinds of jobs in the evolving landscape - same as always.

I just wrote a longer reply to a sibling comment addressing exactly this: https://news.ycombinator.com/item?id=44574895.

[SchemaLoad]

What happens is those companies will end up paying a lot of money for people with no experience when they end up being the only ones available to hire.

Same thing happens in tech in cycles where companies fire all the juniors, and every time theres a boom, they hire them all back at obscene prices.

[echelon]

> The problem is, AI is good enough to replace juniors. That means companies are already cutting positions at that level and some are just itching to ditch intermediates as well once quality improves.

Who needs companies? Those sounds like an unnecessary middleman to me.

[bee_rider]

I don’t see much risk to creatives and workers (hold the number of workers constant and have them all get more productive, and the world is a better place).

But, I do wonder about smaller teams accomplishing more, which the unit of work-doing-people can be much smaller. The org tree can shrink, probably knock out some levels. Maybe a team of 4 that does the work previously done by 10 finds it easier to just have someone directly interface with the customer, rather than needing a layer of project managers and customer service to receive customers messages and distribute them.

I’d be worried if I was… anywhere above an IC really.

[mrandish]

> Maybe a team of 4 that does the work previously done by 10

Interestingly, I heard similar ratios were being discussed (or feared/lamented) in the mid-90s as computer-based non-linear video editing swept the post-production industry and in live TV production as computer-based all-in-one production switcher, effects, titler systems like the Amiga-based Video Toaster disrupted everything. I still remember hearing about how the Toaster was giving unionized TV stations problems because union rules said the switcher guy wasn't allowed to touch the graphics/titling system or the editing system. So three guys would be standing around one chair depending which tab was up on the Video Toaster screen :-).

Having been around the content creation tooling business for so long has given me perspective. Since at least the late 80s it's been one never-ending disruption. And, despite the constant predictions of jobs lost, today there are far more jobs in content creation than there were then. Of course, the job descriptions, types, skills required, divisions of labor and which roles were more or less valuable have never stopped changing and probably won't. Yes, this ends up displacing some people in some roles. Back in the 90s highly-paid senior video engineers with deep expertise in how to wire up and synchronize analog video systems initially laughed at us "young kids" with our "toy" digital video systems. Then they resented needing to call us in to get computer-based gear installed, interfaced and working. Then they resented that us "new kids" were paid less due to being less senior but becoming increasingly essential. Some of them learned the 'newfangled' digital systems while others didn't and opted to retire or do something else.

Over thirty years later, I'm now the highly-compensated "old guy" with once-invaluable expertise that's depreciating by the minute and proven skills at doing increasingly irrelevant things. The particulars change but the theme remains the same, which is why I don't think the shift to integrating AI in content creation will be significantly different. There will be skills that become less valuable and job types that go away and entire industry sectors which get disintermediated but at the same time new kinds of work, industry sectors and business opportunities will emerge in different places and ways. As always, the new job types won't be 1:1 replacements for the old, and this will cause as much angst as it always has. They'll not only look different and have different kinds of trade-offs, they may even have different business and compensation models.

It's Schumpeter's process of "Creative Destruction" and renewal (https://en.wikipedia.org/wiki/Creative_destruction). The destruction part is fast, loud and shocking as what exists today combusts in flame. The creative renewal part tends to be more gradual, unfamiliar and at first may not even seem related. From time to time, even the definition of "the industry" ends up changing along with the jobs and roles. I'm pretty sure the old-school TV station engineers I learned from wouldn't consider a digital nomad working full-time editing 30 second clips on a mobile device for YouTube, Instagram and TikTok influencers as even being "in the TV business."

[minimaxir]

Due to Blender being OSS and scriptable with Python (https://docs.blender.org/api/current/info_quickstart.html), it wouldn't surprise me if a future iteration of genAI/agents works with Blender directly.

[homarp]

There is a Blender MCP server https://blender-mcp.com/

[homarp]

and now there is a second one

https://blender-mcp-psi.vercel.app/ and https://github.com/pranav-deshmukh/blender-mcp

discussion: https://news.ycombinator.com/item?id=44622374

[cultofmetatron]

blender is the python of 3d animation. Its the second best tool for everything. Which btw is not a digg. its really hard to be that level of reliable accross an entire workflow.

[Joel_Mckay]

Maya hasn't changed much in a decade, and that is a wise choice for training/documentation. They focused on the content part which is its value proposition.

Blender has always had the perpetual Beta problem, as many boring core design issues were never really given priority. Fine for developers, but a liability in commercial settings.

Houdini is interesting, but with Blender Geometry-nodes now working it is unclear how another proprietary ecosystem will improve workflows. =3

The Entagma channel covers a lot of Houdini and Blender bleeding-edge features with short lab tutorials:

https://www.youtube.com/@Entagma/videos

[MrScruff]

I've only dabbled with Blender but from what I saw of geometry nodes it's quite a long way away from competing in the same space as Houdini. Houdini's biggest single feature and the thing that allows film studios to use it at complexity and scale is the HDA system and there's no real competition for that.

[const_cast]

I think the main difference between OSS like Blender and the competition is it seems like OSS only gets better. Each version is a little bit better, so projects that were once not competitive catch up. Blender is an obvious example, but also look at Krita, or the entire KDE project. These pieces of software age like fine wine: they get more features, they get faster, and they get more stable.

Closed-source software seems to get... stuck. In the best case. Often, they regress: becoming buggier from version to version with less features. I think of Windows and the entire Microsoft suite of applications.

I think one exception is Gnome. Gnome loves removing features more than they love not implementing popular Wayland protocols.

[Joel_Mckay]

With OSS, unless every add-on is part of the build tree or package repo testing... it quickly becomes broken as the ecosystem constantly evolves.

i.e. of the $3k of community Blender plugins/add-ons we evaluated last year, only around 60% are still functional in stable blender releases. Additionally, many built-in core features like Fracture became broken in 4.x due to API permutations, and getting split into its own module.

In a production environment one must version lock Blender for the project. =3

[const_cast]

IMO it's always a good idea to vendor software. But yes, OSS typically moves fast. It's a tradeoff.

[MrScruff]

Blender is definitely getting better compared to things like Maya but I don't think this argument holds for Houdini & Sidefx.

[Joel_Mckay]

Blender has solved several key challenges, but the learning curve is steep given many tutorials are version specific.

I usually recommend these courses to users when the under $20 sale is active.

They cover a lot of Blenders non-intuitive workflows :3

"Complete Blender Creator: Learn 3D Modelling for Beginners"

https://www.udemy.com/course/blendertutorial/

* Basics of low-poly design in blender

"Blender Animation & Rigging: Bring Your Creations To Life"

https://www.udemy.com/course/blender-animation-rigging/

* Some more practice rigging

"The Ultimate Blender 3D Sculpting Course"

https://www.udemy.com/course/blender-3d-sculpting-course/

* Sculpting, Retopology, and VDM brushes

* basic anatomy

"The Ultimate Blender 3D Simulations, Physics & Particles"

https://www.udemy.com/course/blender-simulations-physics-par...

* Shader/Texture basics

* Geometry node basics

* Boid sprites

* Hair and physics simulation

* Camera FX, and post-render filters

* Instructions on how to export your assets to Unity 3D and Unreal game engines

[brcmthrowaway]

What is HDA?

[raincole]

Houdini is entirely a node-based system and HDAs are just your custom nodes (on contrary to the built-in ones).

At the end of the day, an HDA is just a 'function' you define like what you do in a programming language. A function can call other functions etc. It might sound nothing special if you're a programmer, but Houdini is the only generic DCC that is built around this idea, making it more like a framework than an app.

[Joel_Mckay]

Probably means a Houdini Digital Asset which is similar to a maya reference.

At one time there was a Houdini-Engine Open Mesh Effect plugin, but no idea if that project survived.

Cheers =3

[raincole]

Geometry node is not Houdini (yet) though. The main reason people use Houdini is it's easy to transfer data between modeling (SOP) and physical simulation (DOP) contexts. And they're working on integrating rigging/animation too.

Geometry node is quite a separate thing from what Blender already has. Last time I checked one couldn't even create vertex groups in geometry node, and there was no way to create an armature there either. (Not sure if it changed since I checked though)

[2944015603]

I think you should've be able to access vertex groups from the initial version of geometry nodes. Certainly, at this point, you can dynamically access (and set) arbitrary vertex groups, including loading the names of the vertex groups you want to access from a csv file, or otherwise synthesizing the names using string operations.

[deadfoxygrandpa]

its actually funny, the first version of geometry nodes was actually more powerful in that way. it didnt provide many built in nodes, but the nodes it did provide were arbitrary property setting/getting. the current and newer geometry nodes system based on "fields" is way easier to actually work with but they don't give you raw access to properties. so like in the first version of geometry nodes you could write custom mesh normals just by setting the "normal" attribute with some vectors, but then from blender 3.0 to 4.4 you could no longer set custom normals in geometry nodes and they only added a set normals node in 4.5

[Joel_Mckay]

Indeed, but it is non-obvious how to handle it...

I found the Entagma lab videos very helpful in understanding how to parse geometry with nodes. It is not obvious (unless you already get CLI pipes), but so worth it when things finally work... =3

[Joel_Mckay]

Indeed, Geometry nodes is new, and I have also found documentation on many operations sparse. That being said, it has proven itself very capable.

The mini labs on Entagma's channel were very helpful for parametric surface operations =3

https://www.youtube.com/@Entagma/videos

[johnnyanmac]

That's often a problem with Open Source. Lots of people steering the ship, and it's hard for any de facto captain to really demand a direction. If they do, they may not have the contributors needed to really see it through.

This can be solved with paid contributors, but FOSS organization don't have the most funding out there. So it can be challenging when trying to attract specialized talent.

[Joel_Mckay]

Donated to most projects we find useful, and also tried a few Blender paid plugins.

The other problem is people Cowing anyone that may not agree with their personal opinions how projects should mature. QED: Our karma scores... lol =3

These plugins made Blender usable for a few projects, and I have personally found value in supporting for asset creation:

https://tinynocky.gumroad.com/l/tinyeye

https://sanctus.gumroad.com/l/SLibrary

https://polyhaven.com/plugins/blender

https://artell.gumroad.com/l/auto-rig-pro

https://fbra.gumroad.com/l/Faceit (for Unreal face mocap app with iPhone Pro Lidar)

[egeres]

> Assuming genAI won't deliver and make that world disappear entirely, ofc.

Even if genAI were to kill the cinema/animation back-end industry, I still think 3D software would be needed to continue videogame development, and godot is far behind in modelling/animation capabilities compared to blender. Hopefully they can latch on this in the long term if it happens that AI keeps eating artists lunch.

[tmaly]

My daughter did the Blender donut tutorial and put her 3d animation into the year end art show when she was 10. Having a great community and resources behind the project makes if very easy for students to pick up.

[spookie]

> *: Assuming genAI won't deliver and make that world disappear entirely, ofc.

The current state of things leaves a lot to be desired. I have yet to see a model, or technique that actually tackles real 3D content creation problems, aside from animation of course.

Most solutions are trying to solve things completely orthogonal to what you really need in the real world. They produce bad meshes, if meshes at all, have bad topology or inconsistent point sampling, and are extremely prone to parroting their training data.

It's kind of incredible how far we have come but seriously, to understand a problem space you need to try to use the existing tools. Only then, do new, and useful ideas pop up that solve real issues. But man, nobody in the space seems to have a clue.

Believe me, 3D modelling software is REALLY good at what it does, it is really hard to try and dethrone human and machine working together in these really well thought out pieces of software.