Basic pipeline question coming from houdini background

Hello all!
I wanted to take a look at a possibility of starting to incorporate blender into a studio pipeline, and surprisingly discovered that it doesn’t seem too pipeline friendly.
the main points i’m investigating now are:

  • python expressions in file paths for images (mainly in compositor and video editor)
  • environment variables in paths (found another thread where people say it’s not supported)
  • customizing nodes’ interface (anyone familiar with houdini will understand what i mean)
  • BLENDER_USER_SCRIPTS variable or similar that understand a list of paths instead of just a single path

can anyone tell me if these things are possible.
just knowing that something from that list is possible would already help greatly.

1 Like

Yeah integrating blender in a standard pipeline might be a bit challenging !

You might find interest in that :

and that :

I think in general studios that can afford a pipeline developer use a slightly modified blender version that they compile themselves.

I’m not familiar with houdini, but I guess customizing nodes interface sound the most “probably not” possible thing to do. You can make custom nodetrees in python… You can look at animation node or svertchok addon for an example of that, but that’s probably not what you’re looking for.

For sure blender isn’t build with being standard as a top priority…, It’s getting better with time but they do things their own way for sure. That’s what’s make blender powerful in one hand, but also what makes simple things like that hard to do, and also very long to end up being implemented.

Maybe try to contact other studios using blender and see if they are willing to share some infos.
In France it’s what is happening, a few TD / Supervisors from different studios try to share knowledge and development to try to re-invent the wheel all together.

good luck !


thank you,

this one i have stumbled upon too, made a similar patch, at least that was easy to patch since all that logic is in python.

now this is funny - they have started discussing it back in 2009, made a patch, and still haven’t merged it into release… why i wonder, what are the downsides?

yeah, i’ve just wanted to add an extra string parameter to an existing node (also maybe a button, a couple of python expressions (drivers)…)

indeed, for now i’m still deciding if all that worth the time at all…

1 Like

Yeah, I think these kind of issues is likely to be better addressed now, since there is more and more communication between studios and devs. I don’t think “what are the downsides” is the best way to understand the problem. A lot of patches adds some functionalities that solves a few people issues. Devs needs to balance the choice of including them or not, to not bloat the code for little reasons.
For that they need to fully understand the problem it solves and how useful it is. Here, for someone with standard pipeline/production experience it is obvious, but it might not be for every 3D users and / or developpers. On top of that blender studio have their own way of handling pipeline, which is good and bad at the same time.

Discoverability of that feature might be an issue also.

For that particular patch it’s moving forward at least ! I’m pretty sure it will end up being included and fortunately it’s not always that long for something to be included !

Good luck !

1 Like

speaking only about that patch to enable environment variable expansion in paths - maybe the fact that it was not included is actually a good sign - a sign of a strong design choices made by the core dev team, maybe i’m (so as some other people) trying to force our established views how to solve this particular problem, and not seeing the decision that blender already provides…

i’m trying to look at blender as if it’s an alien that does not follow standard practices, but surely it has to address most of the same problems as industry standard DCCs

yet for now i haven’t found a “clear” blender way of addressing those

thank you!

1 Like

Yeah ! Sometimes it’s the case and they brought interesting new ideas to the table. Sometime it’s just plain non-understanding :smiley: .
For that particular case , sadly I think it’s that they don’t see the point, but as you said they might need to solve that for their own movies.

They use their short films to evaluate blender in production, but this differs a bit from standard production :
One big point is that blender is the main DCC and they don’t have a pipeline that mixes Zbrush, Substance, Nuke, Avid/Premiere…
They also don’t work with different projects and legacy software version in parallel.
Because of that, they tend to be less self aware of issues related to that.

Also, compared to other DCC they don’t depend on the industry. They don’t get a lot of benefit for having blender adopted by many VFX or animation companies especially if they don’t contribute to the code. Therefore they don’t try to be in close competition to other DCC.
Their main goal is to provide 3D creation to the masses, regardless if it’s a big studio or a few individuals. That leads to blender being really cool but not 100% standard.

On the bright side, they share all their production process, so you might dig a bit into how they did their open movies, and how their pipeline is structured.

TBH , I build a small pipeline for animation series and I didn’t feel the need of using environment variables.
Here is a demo of the pipeline :
And a project done with it : Boon & Pimento : Cartoon series made with blender
May I ask how it is useful to you ?

1 Like

I am not from a Houdini background but I have tried it out in the past, and recently downloaded the latest Apprentice NCR version. Anyways in my own short film project I have dealt with a lot of pipeline issues and continue to try to develop my own pipeline of sorts. It’s kind of hacky and cobbled together but it works.

My short film started way back in 2009, it was an attempt to combine Blender and Renderman together when the technology to link them was in it’s infancy. I devoted my whole work to this and ultimately put the project to sleep for a number of years because we tried switching from Blender 2.49 to the 2.5+ builds which broke our entire pipeline.

I resumed work on it 4 years ago and am now using Blender entirely because I just want to get it done and Cycles seems to work very well for the final frames. Plus I just can’t seem to get Renderman to install on Ubuntu 20.04 (no matter what I do). So none the less I press on.

I now use Blender entirely as I said, even have a very tiny renderfarm using CGRU / Afanasy, project management using Kitsu, compositing with Natron and have been expanding my skills with addons I have found that use SVN and asset management. On the asset management side and SVN I have not gotten that fully established simply because my short film has so many linked files and whatnot that re configuring my entire directory structure would break what I have going on… so I have been testing it with new projects and plan on using them for future work. Anyways…

In my own experimentation with Blender and Houdini I have been able to make some Geometry Node stuff and export them to USD and import them to Houdini, so it’s pretty cool to see this work actually. I have NOT experimented with animated geometry yet though but would assume that would too work fine but again I am just getting re-familiar with Houdini after not touching it since like version 12 and even then I had limited experience with it, mostly doing stuff like shaders and some dynamics.

I view Blender and Houdini as compliments to each other. Now that Blender has matured quite a bit since 2009 and even more so since I first started playing with it in 2000 I can say that it’s quite the powerful tool. Will it replace DCC’s like Houdini, Maya, Z-Brush? Nope. Studios have proven and reliable pipelines in place and artists that use these software swear by them and are very good at using them. What I can say though is that Blender would be a nice tool to have in addition to them. It’s not like it’s going to cost more in terms of actually getting Blender, sure there would be costs to pay software engineers and TD’s to get it to play nice in these pipelines but that is the beauty of Blender, you can change the code to bend it to your will and this has been done before so it’s not impossible.

I do agree there could be things done to improve though and someone from a Houdini background could provide some insight into that.

Anyways if you want to read about my work on my short film you can go to this site -

I try to post often as I can but life gets in the way sometimes. Have a read and I would love to keep this thread going regarding this topic.

1 Like

Hey !

That’s interesting ! It’s an impressive amount of work for a short film,

Talking software and pipeline out of context is really difficult since project size and companies call for very different needs and organisation.

For instance using Maya + Houdini + Shotgun + Nuke for a indie short is probably overkill, and chances are that one of your precious team member will end up acting as a pipeline TD instead of doing art stuff.

On the other hand, doing all the VFX of a blockbuster in blender, editing in the VSE, or using google docs as a tracking software in that case doesn’t make sense either.

It’s really a tough balance to find since it’s easy to end up having a very complex structure for what you’re doing. At least when you are in a company it’s always possible to stick to a standard pipeline aligned with other companies doing the same thing.

It’s much more difficult when you are a small studio or individual. At the end of the day, I find that pipeline in general doesn’t make things simpler or easier, but it make bigger projects possible.

1 Like

indeed contribution from big companies is usually small, yet recently more and more internal tech is being opensourced: vdb, opencue, usd, moonray, rv, a bunch of DCC plugins/integrations (in part due to studio’s inability to maintain the codebase themselves), so having similar opensourced blender integrations and plugins would benefit the community.
I think if you want smaller studios/individuals to adopt blender in production - you need to make it easier for them to integrate it in main pipeline.
a fun thing: maya sucks at basically everything, even in animation, i bet blender can do all the individual tasks better, but maya integrates great in pipeline, so great it was the cornerstone of the pipeline (and in some places still is) for years, and recent years it’s replaced in that role by houdini. So, maya sucks in animation, but there are so many free and non-free tools for it that make it usable, and so many animators are taught to work in maya (and a lot are really reluctant to learn new software) - so you have a huge market of animators and tools - in my opinion that is the only thing that keeps maya afloat.
I wouldn’t choose blender as animation software for a studio only because i will have problems finding animators for it.
If blender makes an effort to insert itself anywhere in production pipeline - it will start to grow the same pool of professionals, and that is a big catalysis for studios to invest more into blender.
but that’s turning into a rant…

env variables (or similar concept) makes it easier to make things more “portable”: you can expose certain parameters as env variables to have them controlled by task management software (for wedging for example), or if you have dynamically created environments (like with rez) - you would want to use resolved paths (for ex. something as simple as where’s the base path for rendering)
but i’d prefer to have the ability to drive filepaths with python scripts (got too used to this commodity in houdini) to have more complicated logic there, including custom uri handlers.
Ideally, an artist should not think of paths at all, so we need to make sure everything comes from and goes where it should be automatically.

1 Like

Hello and thanks a lot for your answers !

About blender and companies adoption, yeah it’s kind of a chicken and egg problem.
I think the main goal on the blender foundation side is to at least provide big stuff like USD, UDIM, Alembic support, these have been seriously taken care of just recently in the last few years.

Right now, things more mundane like these env variable support or simple things that would help a lot is in fact much more difficult to bring to the table. Because for that some communication is needed between studios and blender foundation. Since the pipeline they build for their open movie isn’t 100% standard they don’t always see the need.

So which company is going to invest time to provide insight of what they need, how their pipeline work ? And does BF will allow development time to study that closely ?

For software like Maya, they’ve been there because of the industry, and there is a close commercial relation between companies and software developers. For Autodesk , the more companies uses Maya, the better since it means more licences sold. For blender, more users doesn’t always mean more money, and more money isn’t their main goal in the end.
So their POV on that is a bit different, so should be people adopting blender as their main software… People in general clearly see the fact that the software is free to use, but it takes more time to get the implication of being open source, and how it changes the user/software relationship.

In the end BF would prefer fewer companies that invest into blender, by providing code, feedback, funding. But it’s going back to the chicken and egg problem. Fortunately it’s slowly getting better since BF and companies slowly make step toward each other , one step at a time.

I’m pretty sure that the same way BF rely on a few artists to provide feedback, some companies that invest time into blender and who really understand BF vision would probably be well heard.

And about filepath ok I think I get it, probably when working only with blender it’s not that big of a deal to do that in a slightly different way, but when you want to share code or logic between software it might be much more complicated.
Indeed right now you can’t put variables into filepaths, it’s more when you open or render the file the path is set dynamically.

I would bet everything I own against you…that’s how sure I am of the end result.
No, if Maya sucks then Blender sucks 10x more and as a result of that you’ll always find 10x more animators for Maya than for Blender.
To be realistic/pessimistic - I wouldn’t expect a change of the situation before 2026.
Houdini will probably take the animation crown from Maya before that.

If you want to integrate Blender for rendering purposes you’re also 1 to 2 years too early.
Full Hydra/USD and MaterialX compatibility is at least a year if not more away.
As an generalist tool for concept artist and an asset production tool it will fit in every pipeline as long as data goes mostly out and not constantly back and forth.
Everything else might be as painful and costly in terms of time and effort spent than you already know, but occasionally also much much worse and absolutely deadly for the project.
Dying from a thousand paper cuts is a saying that comes to mind sometimes when I work with Blender.

that’s what i’m not quite getting - what is their aim.
if blender is not “on the market” - professionals won’t invest time studying it, and blender will be kinda left out in the “hobby” zone.

maybe i spoke too soon: i definitely know that maya sucks, at animation too, and only clings to the market because of tonns of public and inhouse tools making work in it bearable. (their investment in bifrost doesn’t seem to be paying off)
and about animation in blender i know nothing for now.

there was softimage - it was better than maya for animation not sure if from the beginning, but, let’s say, since a decade before it was bought by autodesk and discontinued, and yet you could still find 10x more animators for maya, as maya back then already assumed the position cornerstone of the pipeline. so having more animators depends firstly on how well the software is adapted by the industry


That was kind of the aim for Blender to Renderman because at least at the time Renderman was so widely used. The code that our team had developed was hacky but it did work at the time, of course the Blender devs had to change everything for 2.50+ so our work became obsolete and the plugin dev stopped working on the new plugin for that series of Blender. Matt Ebb came along and developed the next gen Renderman exporter which was adopted by Brian Savory while he was at Pixar at the time, then Pixar itself undertook development of that addon which exists to this day.

Of course that is just the rendering end of the spectrum, which was our goal.

Now Blender has USD support, can use OpenVDB data, among many other things obviously. I think the USD support was a huge advancement in the possibility of plugging in Blender to an established pipeline, maybe not for final renders but in the modeling and animation departments having USD built into Blender allows one to say model an object or scene in Blender, save it to USD, where Houdini can pick it up seamlessly and go from there. Blender 2.49 did not have this when Project Widow started so all I see is advancements since then. Also keep in mind the industry has in the past decade alone has been more open to releasing code to the public than in the past, OpenEXR was in my opinion the start of that trend, now it’s a standard. now we got USD, OpenVDB, Alembic… these things made a difference to Blender’s adaptability. Had Blender not added some of these I don’t think professionals would even give it a chance, now there are doors open for Blender, I think it’s a mindset that it can’t.

Animation has some ways to go I think, it is getting there. In my short I have a single character rig for my main subject and animation is so slow that it’s hard to really SEE it in real time, no matter what I do or try. Of course all this data I have is converted from Blender 2.49 to Blender 2.79 and then upgraded to Blender 2.93 which is what I am using for final. Things have advanced even further than that since this release. So time will tell.

Now to get into the specifics of Houdini and Blender I have only scratched the surface in my own work, making geometry node scenes, exporting them to USD and then bringing them into Houdini. From there I have to learn more about Houdini itself to really make the most of it but the capability is there. It is really cool to make something in Blender and then switch to Houdini for further work. Since I have the free version I can’t exactly test Houdini to Blender workflow but I imagine it would be possible to do so.

Anyways I am rambling.

1 Like

Bear in mind that just my own POV as a long time blender user : I think they target blender for people using it. In one hand they, for sure, want to make the best software possible, but they probably won’t target the VFX industry hoping for companies to invest in blender. They expect more the opposite : companies to jump in and help blender to become better.
Which is how opensource works in the first place after all…

On the other hand, they don’t want to make hobbyist software either, but probably their want to make sure people or studios that want to build a kind of “open source” pipeline can do it with blender, without the need of third party commercial apps. But of course they are not against blender being used broadly by the industry too.

We are back to the chicken and egg problem, but at least it worked to some extent so far, these last years blender got more and more attention, but maybe not enough to satisfy everyone.

At least in animation it’s working, I see a lot of animation studio moving to blender and a part of that is for using 3D and grease pencil. I think because BF open movies are the closest to animation series or small/mid budget feature-films (animation) the tools tends to fit better.

And yess as @Romanji pointed out , it’s used a lot for modeling or by concept artists, when it’s easy to be out of pipe or at the beginning of the chain.

For complex VFX shots involving a lot of geometry / assets / softwares, and/or 100 Millions $ animation feature films I don’t think blender is there yet.
Part of that is that it’s still hard to integrate into a pipeline, even if it’s getting better. Another reason is that blender don’t scale very well when used with lots of objects or geometry. But it’s getting better by every versions, so we’ll get there eventually.

I don’t think it’s the best way to judge blender’s animation capabilities. Animators animates on whatever software is used on their productions. Since blender productions are still rare it’s natural that there is less animator using it. Also most schools teach Maya… But in general animators moving to blender only struggles for 2 weeks and after that they tend to like it.

1 Like

Sounds weird, in general it’s possible to have a few characters and get real time in workbench engine.
You might need to simplify the rig a bit.
if you look at most of the shot there :
They have two characters, props , sometimes animated sets and play between 20 and 25 fps

When animating generally animators hide irrelevant parts ( characters , sets) to get the animation even faster.
Yes blender isn’t the fastest, but if you dig a bit into it I think it’s very workable.

Worse case scenario you can always have different simplified version of characters like they did on NextGen:
(super interesting conference as a testimony for blender used on a feature film, even if a bit old )

1 Like

Everybody I know that has used Maya hates it, but the abilities and achievements of the software are still there - we all have seen movies done with it and it regularly (almost yearly) gets awards for it, while Blender isn’t really used for feature films besides small productions.
The ratio of successful Maya to Blender productions is 10000+ movies and countless TV shows to maybe 10 if I am slightly hyperbolic.
Blender is unproven in the eye of the outside observer, but also not really able to impress…chicken and egg problem again.
I know how to use C4d, Maya, Blender and I am in the process of learning Houdini. With the exception of the latter one - I love and hate them all, because I know them all.
If you think animation in Maya sucks, working in Blender might change that impression …
At least Maya is bearable :grimacing:
sadly, bearable is the best the 3D industry has to offer when it comes to animation (for now).

True, if you would force me to choose a software for animation between Blender and Softimage XSI - I would take the dead software and I wouldn’t hesitate one second.
Not only am I not convinced of Blenders abilities, I actually truly dislike anything in regards to animation in it, it manages to be worse than C4D.
But these feelings are not just personal preference, Its based on practicality.
I can’t animate if the rig is running at 5 FPS, meanwhile in Maya I get 120+ on the same hardware - unacceptable for me. I can’t stand it if I have to work harder and longer for nothing if I know it could be so much faster.

Stylized and simplistic stuff is not an good example for the ability of the software, I for example have no interest in it.
No animator that has experienced fluent animations of fairly dense and realistic character meshes will ever use Blender and like it if it is 20x slower than what they are used to.
They’ll turn 180° pick up Maya again and trash Blender in any conversation it comes up.
I’ve been in these conversations…
I might be a harsh critic, but IMHO Blender absolutely fails when it comes to VFX (simulations) and animation and without a complete overhaul or rather new build up from the ground systems this will remain to be the case.

Fair enough !
Yeah probably blender won’t work for complex VFX, TBH I never worked on VFX production. And for sure I won’t bet on some Life of Pi VFX done with blender.
It’s quickly obvious that tools (simulation) don’t work very well yet and are fairly limited.

I think it’s important to give context so we are able to understand each other, are you talking about animation tools or just the fact that animating complex mesh gives poor performances ?
Is it an animation tool problem or performance problem then ?
Can’t these performances issue be mitigated by animating lower res meshes, and do simulation stuff in houdini for instance ?

No character animator that I know of was complaining about blender animation tools.
For performances I also heard about very poor ones on Maya because the rig was too heavy (yet again on cartoony animation projects on very well known IP).

For what Barnstorm did on Man in the high castle blender seemed appropriate :
But for more complex VFX I think it’s probably not ready yet.
Blender Bob’s company seems to do a bunch of VFX in blender too, why switching to that if it’s such a no-go ?

Since the OP is about choosing if they want to switch to blender it might be interesting to ask what kind of project they do with it and provide advice according to that. And also what do you think would work and won’t …

Yes I’ll probably agree to them, but what did they try actually ? How they could have tested a production ready character / rig ? Or it was just a quick test ? How could it be a good test if no experienced blender TD was involved in it ?

I heard a lot of people (professionals) talking about blender like that, but it turns out that many don’t even tried the software, it’s just gossips people are passing on or stuff that has been fixed for years. Once I ask for precision I’m always surprised. Since people don’t provide context and try to be as convincing/clever as they don’t know, it’s super easy to spread bad or outdated informations.
In the end, they are probably right but not for the right reasons IMO.

It’s a bit sad to see blender evicted from production because 2/3 TD where pretending to know what they where talking about when having a beer.

In the end I agree with you : for sure animation playback performance isn’t blender strong point. Quite the contrary. I’m just saddened by the lack of balance in comments about blender. Why not show some examples of projects where it worked and some examples of where it won’t ?
Why people feels necessary to go with sentences that implies blender is worse that the plague, without providing a fair context and examples ? How comes people/companies do some projects with it then ?

1 Like

both actually.
There are quite some tools not available in Blender.
Nothing big has happened in animation in the last 8 years since I started using it.

Yes, they can, but the same applies to Maya - stuff easily gets so complicated that working in Maya and Houdini gets painful too - which means they become impossible with Blender.
You might underestimate how often people push to the limit - if Blenders limit is closer - it simply means you have to work harder and longer to get to the same goal, or you simply fail to reach it.
Business owners will look at the maximum they can get out of something - if the general performance or quality is lower with the same amount of investment (or on the same hardware) - they’ll always pick the one with more potential, even if they don’t necessarily need it.
Its a healthy thing to do even if it means being brutal in ones judgement.

They mainly did set extensions - environments and stuff like that. This is one of the areas where Blender will actually work fine, especially today with USD support.

Well, I understand what you say, but again, it’s hard to get what kind of projects you’re referring too since your statements tend to be as if everyone is working on the same things that you have in mind.

I think it’s totally fine to discard blender if it’s not the right tool for the job, but since you’re a long time blender user, I suspect that it’s not all about using it to be sad and enjoy suffering.

Probably blender can find it’s place in some VFX studio for simple but useful tasks ? And that might sparks some communication between studios and BF, adding more VFX oriented tools one at a time.

If the only thing VFX industry as to say about blender is a plain “it’s broken, run away…” I totally understand BF not wanting to be involved either.

Personally I don’t think blender will replace Maya in the VFX at least in the near future, but probably it will eventually be good enough for more and more tasks and projects. And that’s how it will grow even more.

But for that it needs users able to communicate clearly about the good and the bad parts so other can make educated choices, and also raise the awareness among the community. Having blender fanboys isn’t helping, the opposite is quite true either !

1 Like

Yes. But such balance sheet is difficult to establish, to synthesize in current period.
2.8 redesign was ambitious, touching to almost every part of software and thought to correspond to a decade of development.
So, current release of Blender is a kind of Frankenstein creature between 2.79 status and future coherent Blender 4.
In Blender 3, a lot of new stuff is not finished, polished. And a lot of old stuff is still there waiting for its replacement. There are also new stuff that was rushed, or did not reach expectations and will be refactored, again.
And there were some parts that were neglected by 2.8 design.

Constat was done at last 2020 Blender Conference that pipeline integration was difficult.
If you look at videos of Conference, you can see that studios working with blender, often made ugly addons UIs with a lot paths taking a lot of place. The word “pipeline” was mentioned in a lot of talks.
But weirdly, the points asked by @kikomiko were not mentioned as clearly.

UI of paths was simply neglected by 2.8 UI redesign proposal.
Patch mentioned by @sozap happened in 2009, during 2.5 refactor, it was lost in the mass.
Then, a comment revived it in 2020, 2 years, after 2.8 codequest workshop.
And it was also expressed in middle of a ton of feedback on 2.8 series.

As developers, working in a studio with multiples machines, I think that they don’t need explanations to interest of environment variables in paths. They simply have other priorities.

They want to improve robustness of work done in Blender, before making it communicate with other stuff.
In all points, a Blender 3 is more powerful than a Blender 2.79. But that is not always as fluent or coherent, because of “Frankenstein” status.
Desire to focus on satisfaction of old users was expressed in Ton’s keynotes of conference.
That is not because they are not open minded.
That is just because of context of current “in-between different designs” status of Blender.
That will be easier to work on communication with other software, when work in Blender will be fluent.
When viewport compositor will complete the package of Grease Pencil + EEVEE (that is making strength of Blender for storyboarding, previz and 2D/3D mix) ; more people will be interesting in testing pipeline integration solutions.

Last autumn, there was a workshop about animation tools. And in article of code blog, Sybren announced that there will be experiments done, about new animation tools, in upcoming year.
There was also a beginning of work to bring simulations to geometry nodes, with a new caching workflow.
Support of Vulkan by EEVEE next and Workbench should improve a lot abilities of viewport to handle heavy scenes.

So, it seems that some animation issues will be solved in a not so far future.

It is always possible to do better.
Complaints about old stuff are legitimate. It is planed to replace it.
Complaints about new stuff may be legitimate, too. It is planed to complete it by other new stuff or polished, later, or it neglected an aspect.

So, satisfaction is something really subjective.
If you are just one individual who don’t have time to modify UI or create a custom build ; and if you are a studio with a Blender dev employed : it is obvious that you will not be blocked same things and don’t have same expectations.
The “good enough” threshold is not the same for everybody.

So, I think that @kikomiko had correct approach by asking questions on specific points.
UI of file paths did not evolve, recently, is very limited and not common to all uses (path for render output has some specificity not handled by path for source of image datablock).
Blender Studio is working with a separated asset manager. And most of studios using blender are building their own workaround.

Customization of a node aspect is limited to a custom color and a custom label.
There is no custom shape or custom icon.
About varying shape, we are limited to opening/closing the node, hiding/revealing unused sockets to modify its height and increasing its width.

1 Like