🔥 Aujourd'hui, sortie de #Avatar3 de #JamesCameron ; occasion de rappeler quelques chiffres #Wikipédia sur Avatar (2009).
ℹ️ Le rendu a reposé sur GNU/#Linux : 4 000 serveurs, 35 000 cœurs, 104 To de RAM, 3 Po de stockage.
💡 GNU/Linux est massivement utilisé dans ce domaine pour sa stabilité et sa capacité à exploiter au max le potentiel du matériel.
#️⃣ #OpenSource #LogicielsLibres #FOSS #FLOSS #FreeSoftware #BlockBuster #BlockBusters #RenderFarm #RenderFarms #FX #SpecialEffects #3Deffects
#RenderFarm
Flamenco v3.8 was just released! Some of the release highlights:
- Settings editor in the web interface.
- Multiple jobs, tasks, and workers can can be selected simultaneously, making mass-requeueing and worker management simpler.
- Compatibility with Blender 5.0
- Progress bars on jobs & tasks, to show how far along they are.
- ... and more!
Flamenco v3.8 was just released! Some of the release highlights:
- Settings editor in the web interface.
- Multiple jobs, tasks, and workers can can be selected simultaneously, making mass-requeueing and worker management simpler.
- Compatibility with Blender 5.0
- Progress bars on jobs & tasks, to show how far along they are.
- ... and more!
—Asking for Advice (Rendering/Renderfarms):
Hey, I never used a render farm before.
Now considering for a recent project.
There are a lot of options…BlenderGrid, Fox Render, Renderjuice, …
Even free (!?) ones like SheepIt.
Am not very techy in this regard. So looking for an easy to use solution.
(Blender 4.2.9 LTS, Cyles, Animation, several rigged characters, indoor/outdoor scenes)
Anyone got Tipps, Recommendations?
😳 thank you.
Nach geraumer Zeit wollte ich heute mal wieder mit #Blender etwas rendern. Da meine Hardware eher schmalbrüstig ist, habe ich vor Jahren mal ein paar Dollar für zwei #Renderfarm-Dienste ausgegeben. Mittlerweile sind sie beide nicht mehr verfügbar. #Seufz
Also dann: Energiemodus => Leistung!
Flamenco v3.7 was just released!
- Load job compiler scripts on demand
- FFmpeg upgraded to 7.0
- Fix issue with pausing jobs
- ... and other goodies!
https://flamenco.blender.org/download/
Flamenco is Blender's render farm system.
Flamenco v3.7 was just released!
- Load job compiler scripts on demand
- FFmpeg upgraded to 7.0
- Fix issue with pausing jobs
- ... and other goodies!
https://flamenco.blender.org/download/
Flamenco is Blender's render farm system.
#Beszel looks like something I‘d like to try for our #renderfarm. Deadline is a bit inadeqaute for hardware monitoring (no temperature, no GPU stats). And you can only run so many terminal windows with htop… https://fosstodon.org/@shollyethan/113741850938551727
Flamenco v3.6 is released!
It adds support for pausing jobs, submitting files from Blender via Python, and more. And of course a bunch of bugfixes.
https://flamenco.blender.org/download/
Flamenco is Blender's render farm software. Free & Open Source, of course!
I've been working on refactoring Flamenco, Blender's render farm system. I'm replacing GORM (an Object Relational Mapper) with sqlc, and it's _so_ good.
Most importantly: I'm more in control, and it's easier to refactor & optimise things.
The more I work with sqlc, the more it shows it was the right way to go.
Check https://sqlc.dev/ 😻
Cinelerra-GG is one of the few video editing programs that allows background rendering over a network with several connected computers. Rendering is memory and CPU intensive and this feature can significantly speed up rendering by distributing the rendering task across multiple computers.
In order to make the next VR cartoon, I'm going to have to render nearly ten thousand 360-degree 16 megapixel images.
Which might be fine if it was mono-vision, but if you want stereo-vision you have to do tricky things with translation of the cameras for each pixel-column in the output image or you'd be cross-eyed when you turned around.
So what you have to do is: build a rig of about 60 1-pixel-wide cameras pointing at right-angles to each other and rotated x/4096th of a circle, and scan that rig over the time-frozen scene as they rotate to fill each of 30 columns in the output image [ https://github.com/revpriest/godotPanoRenderer ]
And the trouble with that, is it takes like two or three minutes per frame. So doing ten thousand of them takes two weeks or so.
Two weeks of render!? Not ideal. Some of the future cartoons are supposed to be topical and done in a weekend.
But what if you could just rent a hundred computers overnight from Amazon? Get them to do a hundred frames each.
Well you can indeed rent the machines, but can I install my software on it and run it on a computer that doesn't have a monitor plugged in and I can't see the windows?
Proof of concept is done: Yes I can!
We install the Nvida drivers, and then X11lib, and vulkan libs, then run startx in one console window, while in another I do the display-config by prefixing the launch of the built app with "DISPLAY=:0.0 " in another console.
A few changes to my code to allow it to accept command-line arguments telling it which frames to render, and my Godot app can run on a remote machine and save the rendered files to disk.
Next up is how to scale that, so I can run it on 100 spot machines overnight. They will need some sort of coordinator scripts I guess.
And the particle effects are going to be annoying since they are random, and will thus be different on each machine. Hopefully I can fix their seeds somehow.
But, the good news: I have successfully rendered an image on a remote amazon machine. Being able to do the render on many machines at once is not going to be a blocker.
The show will go on!
Wasted hours on the #RenderFarm because you forgot to check your sampling settings? Hop on the bike to make up for it!
Simulated a candle flame at 2 hours per frame because your sim container was 10 meters wide? You better get pedalling!
The vfx industry will never be the same…
Debugging in production on the #renderfarm works great if your scripts react to a certain DEBUG environment variable though. Use it to set logging to debug level or do some extra prints. If your renderfarm allows you to edit the environment of jobs already on the farm (#Deadline does) you don’t even have to submit test jobs. Just requeue some existing frames.
I will release Flamenco 3.5 soon. Please give the beta release a try, and report any bugs, especially regressions since 3.4!
Flamenco v3.5-alpha0 is now available for download at https://flamenco.blender.org/download/#go-experimental
This includes the built-in MQTT client for connecting your render farm to to HomeAssistant @homeassistant
Please note this is an experimental release. Report any bugs you find, but don't make your production depend on it quite yet!
For the other changes, see https://projects.blender.org/studio/flamenco/src/branch/main/CHANGELOG.md
This year Flamenco will be part of the Blender GSoC again! So if you want to join in development and get mentored by me, this is a nice opportunity.
There are already a bunch of idea suggestions at https://developer.blender.org/docs/programs/gsoc/ideas/#flamenco
Come over and have a chat at https://blender.chat/channel/flamenco !
I just released Blender Asset Tracer v1.16
It fixes an issue with Flamenco/Shaman for the upcoming Flamenco v3.3 release. And it has a nice new feature to get more strings from blender data.
It can be downloaded from https://pypi.org/project/blender-asset-tracer/1.16/
Friday at #bcon23 I'll host a hands-on class on our render farm software Flamenco. Get it running on your own laptop, right there on the spot!
Join me in the classroom on Friday 12:00!
The linked post, for easy access:
Flamenco v3.3-beta1 has been released!
It's an experimental release, so be careful. Any testing you can do would be much appreciated.
Downloads are available at https://flamenco.blender.org/download/#go-experimental
It has some nice new features, like clustering of workers, automatic detection of the project's top-level directory, and many more. The download page links to the entire change log.
Flamenco is the render farm created by Blender Studio.
Client Info
Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst