Welcome to the Blender-Community dear Wenzel!
That’s so cool that you respond so fast. After we all worried mitsuba could maybe only be a short one man project you came and say " it’s definitely here to stay for a while" - very cool statement!
Your work looks very promising and professional.
You could be a big help for the blender developers(and artists). You can find the developers in irc server: freenode #blendercoders
“I’d be grateful if you could send me a copy so that I can go bug-hunting” - man you’re to kind, we are grateful that you found us and for your nice and detailed explanation.
In-terms of stability, speed, and features Aqsis is not even in the ballpark. I’m sorry but its true. i respect what the Aqsis team has done so far but the project has been around for quite some time and its still not production ready or proven. If I’m going renderman I’d rather pay for an Air or 3Delight license.
Personally the complexity is not my issue with Aqsis, there are lots of tools that make things easier, from shaderman to RIBMOSAIC among others. Aqsis is featureful but its not very stable or fast and the team behind it aren’t production focused, which is fine. What I want to see is an open source renderer with real film and animation production in mind from the outset. Like I said I thought Lux was going to have that focus but I just noticed their mission statement seems to have changed slightly from being a PBRT renderer for production to being a PBRT renderer for artists. If someone could take Lux add a shading language or material api, fast production quality motion blur/dof, a flexible file format that can be easily parsed, it would be perfect (well barring other things that make a production ready renderer). A shading language is all its missing. Well there are other things missing but those can be worked on.
I share your feelings on this one, there are a couple of reasons why IMHO there are no “production focused” FS/OSS renderers alive.
One is that the developers get aggressively recruited, such as George Harker who got recruited by Pixar, leading to the death of Pixie. Or atleast put Pixie in some kind of zombie state
Interesting quote from an old CGtalk thread, where a well known Renderman guru who’s worked on alot of feature film productions said the following:
One other big reason is that the userbase itself of blender either doesn’t understand the importance of a flexible shading language, render-time displacements, stochastic motion blur and depth of field, etc. Or it’s not just relevant to them and what they use the software for. So these kind of things never get enough requests so the developers don’t add them.
If LetteRip manages to get that subscription based business going things might start to move in that direction though.
But, Mitsuba seems like a good platform on witch to build a production quality raytracer (ala Arnold, V-Ray, mental ray), if the author is willing to either fork it, or collaborate with developers interesting to make the relevant changes.
Lux and Yafaray I suspect has already deviated too far from this a long time ago, whereas Mitsuba is in a pretty early stage.
And why is this the userbase’s problem, Hell they can understand the importance of shading language, render-time displacement, stochastic motion blur or depth of field but understanding the importance of these things is not what brings features into Blenders. Code does and that is dependent on developers not the general userbase.
Its also dangerous to assume that just because someone is a developer they can easily switch hats from one day and to the next and work on multiple areas of Blender. So making feature requests when there isn’t anyone to work these things is also futile.
Yes but there’s still a connection between popular demand and what get’s focused on. In this case it hasn’t been popular demand at all.
But to be fair in blenders case the internal render would require almost a total rewrite for stochastic rasterization and there aren’t enough developers for that. It requires several full-time developers.
Had there been a push from the users very early on it would perhaps have been possibly but most likely it would even at the early stage be too big of an effort, and everything else in blender would have halted.
So in Blenders case this doesn’t completely apply.
However, for NEW renderer projects this does apply, even for FS/OSS where it’s mostly what the developer finds interesting/fun to add the ends upp getting added, if there’s an overwhelming push among the user base for something it would most certainly affect the goals of the developer.
But FS/OSS renderers are about the last thing film production studios look for, since they require at the very least a guarantee of support, to be able to get bugs squashed instantly during the deadline of a project, so they haven’t made their voices heard and haven’t been part of the userbase. Thus the userbase mainly consist of allround CG artists that mainly produce still images and the development of the software reflects this.
I didn’t mean it’s the fault of the existing userbase, just that the part of the userbase that wants film production oriented features are a very small minority.
Paid support and/or subscription for FS/OSS projects would change this entirely.
And there seems to be a push for this from people in and around the Blender Foundation now (current discussion on the mailing lists).
Nope, at least this is not my observation. What gets worked on are demands or feature requests that happen to dovetail with what ever a developer is working on and has more importantly the knowledge to work on.
For example if someone is working on sculpting tools and you happen to make a feature related to that area of development it has a high chance if been picked up and worked even if you were the only person to suggest such a feature. If 100000 people beg and plaid for a feature its not going to make a difference if there is no coder with an inclination to spend there free time working on that area of Blender.
Funny that because all the recent improvements to the render engine have targeted that very crowd. AAO, indirect lighting etc. BI serves animators better than stills people.
I agree that Aqsis isn’t 3Delight or Air, however I disagree with the rest of it. Aqsis isn’t production proven because no one has taken the chance and tried it out, well until now, hopefully in a few months there will be a short animated film which shows it’s capabilities, and yes I know a short film is a little short of a full production, but it does go a long way in terms of testing stability, speed, and features.
The team (Myself, Animate1978, and WhiteRabbit) give feedback to the developers of Aqsis which helps them with fixing the speed and stability issues, as well as adding required features, if they don’t have this then how can they produce a production capable rendering engine, as far as I know there isn’t a ‘standard’ list that fits all, there needs to be a cycle, and to be honest no one has bothered to offer Aqsis that yet.
Shaderman and RIBMOSAIC do make it easier, we are making use of these tools in the production, but I still stand by in my beliefs that even with those two tools, using a rendering engine like Aqsis, with custom shaders etcetera in a pipeline is still too complex for most users, and to be honest is probably overkill considering the Blender Internal will do most things.
If you look at the recent builds of Aqsis you’ll see huge improvements to the speed, stability and features, I’m not sure if you’ve checked lately but there are some fairly big improvements which have directly resulted from the feedback the whole team has provided.
This is how a production renderer comes to fruition through feedback, bug reports and talks with the developers about issues and needs that arise from real world productions, be those productions short animations, full length films and/ or visual FX productions.
I think that many of the reactions you can see in threads like this have more to do with the fact that people do not realise that, while writing a raytracer can take a few months, taking it to a production-ready state and building a decent community around it can take several years, and an awful lot of work from many people.
When YafRay released the 0.0.9, we did not have a competitive engine, we did not have any developers, but we had a community and a brand, and the work done during 5 years, and that was enough to keep on going.
Much respect to Alvaro, Yafray has been accessible to Blender since at least my 2.3 guide, that’s pretty impressive indeed. It will definitely take time for the new guys to prove themselves, that’s for sure. Good luck and happy coding!
Yafaray is in my opinion the last chance for Blender to be taken serious in the CG area.
Blender Internal is not capable to compete with any modern engine and as it looks like
it is also not on the plan (maybe also not possible) to bring Blender Internal onto the
same level as other engines.
After all Blender has a great platform and most professional systems I used
Maya, 3DMax, and others utilize the same approach.
A host application and an external render engine.
Which Hollywood movie is made with Maya Internal Render engine …
Alvaro
when I entered Blender years ago, Yafray was the only real option for GI.
In the recent years because of Blenders development access to external engines
and exporters were an issue.
But I am sure when Yafaray has the exporter running in a usable state that Yafaray
will again be used more - in particular because it offers so much more than Yafray did.
Dont sound so frustrated you have a strong following base - people like me were just
not really able to use Yafaray because of Blenders changing nature.
Much of the criticism towards Blender Internal (hereafter BI ) seems to be about speed for animation … I gotta represent the still makers; pardon please … and it is prossional use, as I make stills for print which I have sold, just not very vigorously.
for stills, I find BI’s performance to suit my needs FAR better than any external engine I’ve tried. It can render a 300 dpi 24 inch by 36 inch image in reasonable time (as tiles, mind you … 24 - 48 hrs) an an old-ass AMD XP 1800+ (single core) with 2G RAM and ATI x700 (256M) with slow hard drives …
but THEN those Blender Genertated files in OpenEXR multi-layer format can be taken back into the Compositor for adjustments limited mostly to how I can figure to use the tools available … sure, it crashes sometimes when you try crazy stuff, but show me a renderer that NEVER crashes … you’ve got to optimize your scenes, materials, render layers and passes, to make the fastest-possible renders and most-adjustable outputs …
THAT’S where Blender Internal WINS as I see it. The files it can generate are FULLY useable in any other part of Blender without translation or (usually) error … there are, of course, no absolutes …
For photo-real scenes, Blender’s output would have to be SERIOUSLY modified to even come close to matching any un-biased render engine sayyyy LuxRender … and then it would only be an approximation.
But the flexibility of Blender Internal WITHIN Blender itself is un-matched by ANY external renderer (except maybe properly-used RenderMan language)
Hope I’m not too ranty, but if the userbase, as hypothesized, is mostly still makers … one of us had to explain our love for Blender Internal. =) =)
I am a still maker and BI cannot deliver what my area needs.
Blender Internal is coded for speed for rendering of animation and is based on what the movie projects required and not what many object render people need.
If your renderings are faster in Blender then maybe that has something todo with what you also render.
In terms of raytracing performance VRay runs circles around Blender.
Very smooth glossy materials are nearly impossible with Blender since the sample rate would have to be set high and thus slowing down the render time significantly.
Also why do I read recently always about Lux vs Blender when speed is an issue.
Those are two very different approaches.
Why not Blender vsYafaray or VRay as both can be set to be biased.
BTW Yafarays raytraced AO is way faster then Blenders AO.
I would also say that this is not really criticism but simply a realistic assessment.
Ton would like BI to be better but with the way how the software is developed this
is just not such an easy task and thus put onto the lower list of urgent projects.
Fair enough =D
I postulate that you are doing architectural or at least photo-real stills. Yes??
Myself, I do more abstract type stuff, surreal, looks kinda real (sometimes) but you know it can’t be, weird stuff … for me, BI is the most flexible option due to the way I can combine the different passes and suchnot … re-rendering a single pass or layer is A LOT faster than re-rendering a whole scene.
I’m also working on a resume to break into the astronomy / sci-fi print image market, and for me, again, BI is much more flexible and speedier for that solution.
Of course, If I want realistic caustics or diffuse-inter-reflection, I turn to LuxRender, Yafaray, or POVRay and composite that layer in best I can …
Actually I do jewelry, Industrial/Prouct Design, Graphic Design(packaging), and everything else related.
I agree with you on one point very much - Blenders platform is terrific.
The way how material, rendering, and compositor work together is amazing.
Plus in addition you have a sequencer as well.
I just wish the Render part would be as powerful because this would turn Blender into a system which could satisfy many needs and not be limited to what it can do and what not.
… you’ve got to optimize your scenes, materials, render layers and passes, to make the fastest-possible renders and most-adjustable outputs …