Unbiased Rendering - over rated?
Dondec
Posts: 243
I’ve been reading about Octane, Luxus, IRay and other new and alternative rendering solutions. Very simply: I’m wondering if people (i.e., "you") are really liking these technologies. No disrespect to the aficionados, but as I’m getting older I've realized how much time I've spent researching, buying and learning new rendering tools, as opposed to getting on with my original goal of “making cool rendered pictures”.
Ok, yeah I’m interested in finding good and better tools (which is why I’m now using Carrara to begin with). But well… lets get started: I don’t seem to be particularly attracted to the unbiased render technology. Yes, the unbiased renders do have a certain look about them, and in some cases can look very realistic, but on the other hand, from what I’ve seen they’re not particularly attractive. Makes me wonder: Has the “realism” thing become a bit over blown? I can't find any enthusiasm for unbiased rendering strategy for animations either.
Sorry. Maybe I've seen too many unbiased, no post work images. Anyway... love to hear your comments.
- Don
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
The unbiased renderers are very good.
But, it's kind of a capitalist/purist differentiation in my opinion.
In pro pipelines, it's a time-money-results thing. I'm a capitalist.
For stills, as long as the CG application has a good renderer, I can make up for any inadequacies in Photoshop very quickly. It's simply not worth the time/expense/learning curve for many of the shops I work with or within my studio.
For animations, there are some other production factors to consider and it is less clear cut. That said, most of the 'big name' CG solutions have very good renderers. What's more, and even more important, they export all of the g-buffer info anyhow so animations can be modified in AE or NUKE as you would in Photoshop without re-render time. Producers are notorious tweakers and there aren't enough hours in the day to keep re-rendering lighting and density changes.
G-buffers rule.
Let the firestorm begin... :D
There are no "bad" render engines. There is only the render engine that works best for you in the render you need to make.
It depends on:
-What you want to use the engine for
-How much work you're willing to do to make it work
-How much time you can allow each render to run
For me, as much as I like the architecture renders I've seen from Octane and Lux, I've seen very few good renders from them with humans. (I have the same problem with Carrara, actually, even if it isn't unbiased.) It can be done, but it takes mastery of their shaders, so there are just as many bad Reality/Octane renders as bad 3Delight ones.
The same is true of Iray, except that it is much faster than Lux (I don't have Octane so I'm not sure if it's faster than that one).
I'm currently using Iray more now because it is fast enough that I can reasonably fit it into my product workflows, and it's built into DS, which also makes things easier for me. I still use 3Delight for a lot of images too, because it's still much faster, and when I want a certain more heightened/stylized look it's still great.
No. It is just another avenue, we all express our self in the way we want ...
I have no idea what that even means, are you saying that you did not see "too many" biased, no post work images ? :)
A physical artist works with a variety of tools, A Digital Artist works with a variety of tools.
Render engines are the digital Artists "paint box" Some like to use one style of paint brush, some like to use another.
When offered a new "brush" to try out a large percentage will experiment to see if it will fit into their work flow and to see if it will improve their output.
Some on the other hand will just ignore it and prefer to use the tools they are familiar with, and are happy using.
It's horse for courses.
You are seeing a lot of unbiased renders at the moment because a new brush or tool has become available at no cost to the artist. It still has the "Ooh Shiny" appeal of any new toy, so you are seeing a lot of experimentation using it.
And because it is new, everyone started off on the same footing. There are no true "experts" yet, although some are getting to grips with it quicker than others, as always happens in any group of students.
Well said!
There is a saying from the Old West, "Beware the man who carries one gun."
Being super proficient with what you have is always better then being mediocre with several.
(note to self: practice what you preach)
Thank you. I guess I should clarify. Not looking for philosophy or justification of using different render engines. I’m hoping to get practical comparisons and trade-offs from folks who’ve actually used both.
Maybe I should state I'm more in the MarkB “G-buffer” camp at the moment, meaning “making pictures” from Carrara’s Multi-Pass renderer in PS or After Effects. What’s neat about this strategy is you can easily improve on the default render, you get more than one camera view at a time (just create a few camera keys on the timeline), AND the compositing setup you create once in AE generally works for the entire set of camera views, and all render sizes done later. Additionally the same compositing setup (layers, opacities, mask arrangement) are easily transportable to PS if you need larger format. I hope the unbiased camp can relate to what I’m saying, anyway that’s the way I work now. I’m iterating toward the goal, sure, but seems pretty efficient.
Thanks SickleYield, this is the first, non gung-ho comment I’ve read about unbiased rendering, and it makes sense. I spent about a week trying to puzzle changes to after market skin shaders I purchased and finally came up with my own strategy. I see I’d have to puzzle that all out again, going unbiased.
Probably not the only thing I’d need to puzzle out… I bet.
-Don
Mastery of lighting is also super important to getting good renders - but that's true in any engine.
However, the carpenter who is super proficient with a hammer probably won't build as good a house as one who also knows how to use a screwdriver. Sometimes being super proficient with one tool creates a tendency to try to do everything with that tool, which may or may not lead to a better result than using a tool you aren't as proficient with, but which is better suited to the job at hand.
However, the carpenter who is super proficient with a hammer probably won't build as good a house as one who also knows how to use a screwdriver. Sometimes being super proficient with one tool creates a tendency to try to do everything with that tool, which may or may not lead to a better result than using a tool you aren't as proficient with, but which is better suited to the job at hand.
But if I have a hammer, the world is my nail! :-D
Thanks another important data point... I didn't know that. Even though I don't get into DS much, I thought Iray was WAY faster that 3Delight. Helps reorient my perspective. Thanks
- Don
Thanks another important data point... I didn't know that. Even though I don't get into DS much, I thought Iray was WAY faster that 3Delight. Helps reorient my perspective. Thanks
- Don
Not in a million years, lol. 3Delight in DAZ Studio, when combined with UberSurface shaders set to Occlusion and 128 samples on any transparent object, plus Advanced Ambient lighting instead of UberEnvironment, is incredibly fast. I can render a 2000x3000 scene in 15-20 minutes with a full set and two characters with hair.
Iray is faster only in very specific situations - those that involve godrays and refraction in particular.
This scene cannot be mistaken for an Iray render; but it took only 15 minutes.
http://sickleyield.deviantart.com/art/Sneak-Preview-Body-Studs-G2-521586090
The goal of many (but not all) people is realism in their renders and unbiased renderers produce - in general - more realistic renders. Those that provide an interactive preview, such as iRay on a suitable system, and Octane, allow you to quickly see a rough of the final lighting and materials, which makes tweaking more of a pleasure than a long-winded, iterative cycle. So there is a lot to be excited about unbiased renderers. That is not to say that biased renderers don't have a place, they still do and can produce excellent images, or you can mix and match between different renders of the same scene - for example to include Carrara hair which isn't supported by any unbiased renderer.
But just as a photograph is a good representation of reality, there can still be a good photo and a bad photo, so too with unbiased renders!
As for animation, I am of the view that the quality of the animation itself takes precedence over the renderer used - if the movement isn't right, then it will still look wrong no matter how good the render.
Honestly I don't think you'll find much of that here. Clearly I think most of the users here are in the hobbyist category, and as such do this 3D stuff or enjoyment. And for many that includes playing with the latest and greatest software because it's fun. That's not a bad thing, it's just how they get their enjoyment. And as chohole mentioned, there's a lot of "ooo, shiny !!!" involved.
Unbiased renders look cool, and watching them rendering real quick in real time fulfills the need for "ooo, shiny !!". And being able to get decent realism with little work increases the enjoyment.
So don't expect that people here will be doing detailed analysis and comparison of the resulting renders to fulfill a certain need to produce a certain quality for a demanding professional project.
And as I mentioned here recently, comparing the output of renderers requires a lot of work to make sure you're comparing apples with apples, and it also requires a lot of skill. Because when you think about it, what makes one renderer "better" than another? If you ask 10 people here, you'll get 12 answers. Some will say "realism", but when you ask what that means, you'll get another 15 answers.
I'm sure this doesn't answer your question. But maybe a better approach is deciding exactly what you need in a renderer, and finding out which can meet that need. Because, as I mentioned, you may be able to meet that need with just improving your skills with your existing Carrara (or whatever) renderer.
Thanks all. At first I thought unbiased was absolutely the way to go, but as I looked into it I started seeing “hidden costs”, in terms of additional render time, learning curves, shader tweaking, and addition expense items like faster GPUs (hardware) and filters to de-pixelate unbiased artifacts (software).
Now I’m swinging in the other direction. Unbiased, I’m starting to realize, has a place: it can achieve a higher level of realism, but at a definite increase in cost in terms of Time (except for maybe godRays and refraction, thx SickleYield) and perhaps money. I’m not hearing anybody who’s used both say, their going unbiased and never coming back. Point taken!
Remaining questions then:
• How do you classify render time when using unbiased (I know they can go on forever, what do you look for, a superficial “it looks good enough” or you getting in there with a magnifier or something).
• How long does an acceptable quality 720p take (ballpark estimate, all related to animation rendering)
• If I double the size of an unbiased render, does rendering take approximately x4 longer as it does with biased (double size is 4x the area)
• This is the big one: Animations and network rendering… possible for hobbyists… or not (I have good but certainly not high end GPUs)
Thanks again, appreciate all the comments…
- Don
I can't answer your questions about animation, because it's not my area.
I'm pretty sure you're correct about size vs. time.
On quality, I want it not to look grainy ("bad" is blurry with biased, grainy with unbiased, usually). How long it takes to get there depends on your hardware. I invested in two GTX 980's so I could do unbiased card-based renders. I can do most promo-size renders (that's 1000x1300 pixels) in around half an hour, if it doesn't require dust effects or other big laggers.
The answers to all of those questions are unanswerable. Except for "it depends". Render times depend on how the lights are set up, how the textures are configured, and on and on. Just like you can get the same Carrara scene to render in 12 seconds with one setting, and 12 hours with a slightly different setting.
As far as animations, again it depends. If you don't mind waiting hours or even days for your renders to be finished, then anything will do. But typically, if you're doing animations in the real world, and you're a one-man-band, you'll probably want to do what professionals were doing for decades. And that is to simulate realistic renders, but with reasonable render times, using a standard renderer, but with some specialized lighting and texturing and rendering skills.
The best answer to your question is it look at the animation you want to complete, decide how many seconds or minutes the completed animation will run, then assume something like 24 frames need to be rendered for each second of animation, and do the mulitplication.
So a 10 second animation will require 240 frames to be rendered. And if each frame takes only 1 minute, that's a total of 4 hours or rendering. Realistically, I tend to shoot for maybe 3 minutes per frame of render time, just cuz I find that it takes about that time to get a reasonably realistic result from a standard renderer. So that's 12 hours to render a 10 second animation.
Now, is that acceptable to you? Because if it's not, then you certainly aren't going to like a render that is gonna take 30 minutes per frame.
Do the math. In general, no matter what renderer you use, the "better" the result, the longer it takes. If you're happy with what I call an 80-90% solution in terms of believability, then you can do it with a standard renderer once you learn the skills and techniques. But if you need to clinch that final few percent of believability, and can stand to wait for days to tie up your machine with a render, then you might want to consider using a fancy renderer.
And if you're just playing with all of this, then do whatever is fun. Try it out. Can't hurt. :) :)
Many un-biased renderers will take as many sample as you'll let them as they resolve the image. It is called physically accurate, but it is still a simulation based on a computer model. and since computers, even todays super computers, have finite memory and processing, short-cuts are taken. Un-Biased renderers "sample" the lighting, but it is impossible to calculate every little virtual photon as it interacts with every polygon and surface property, so it will maybe calculate every tenth photon or polygon (have to look it up later) and extrapolate. The more samples that you let it run, the more the image get filled in, and less grainy and sparkly it gets.
Jeremy Birn's book, Digital Lighting and Rendering, has some very approachable and easy to understand explanations on how a biased renderer and an unbiased renderer works. He also describes the pros and cons.
I highly recommend this book. It is written by a Pro, and some of it is written from that point of view, with the assumption that the reader is headed in that direction. However it is also very approachable for the independent artist or hobbyist/artist. It is also fairly platform and software agnostic, although you won't see references to volumetric lighting or HDRIs called Uber-Anything-Lights™.
;-)
Great info SickleYield. Thank you once again.
Thank you very much EP, Hmmm... I think I have one of his books somewhere, anyway... will check it out.
- Don
Thank you very much EP, Hmmm... I think I have one of his books somewhere, anyway... will check it out.
- Don
I think I have the third edition, where the un-biased rendering information was added.
Don,
This is a two part response
Part 1
I think a lot of people truly misunderstand the necessary value of realism in CG rendering. It's about VALUE. I will explain.
Realism in a technical sense is what its always been about. Even the biased engines we've worked with for decades had "realism" in mind when they were originally developed. Compromises were obviously made in the name of speed-up sure, but make no mistake, most biased engines are only biased because unbiased simply wasnt feasible until recently.
"Realism" provides a self consistent "metric" by which any content can be appraised at any point in time without hesitation or external references.
It is important in life to have Metrics by which we can measure the quality of competing products. We need a way of distinguishing the "expensive" from the "cheap."
"Do you like this image?" is too broad of a question, and most often relates to whether the viewer agrees or disagrees with the overall subject matter of the piece. Lets imagine it's an image of bunny rabbits eating carrots. If the viewer happens to find animals and rabbits in particular interesting, they will probably say they "like" this image. If they hate animals and in particular hate rabbits, then they will vote my image down. I realize that a lot of people may disagree with the subject matter of a given piece, so I don't really care about their opinion on that issue. But when it comes to technical considerations like "does the fur on this rabbit look real to you, does that carrot look real to you, do those rabbit droppings look real to you?" are questions any viewer can answer, even if they happen to hate the subject matter itself. I like that anyone can participate in such a discussion.
The viewer then responds "I guess I just feel really bad for the carrot." At which point the artist responds, "don't feel too bad, it wasn't a real carrot to begin with." Viewer, "Really, it looks so real!, thank heavens."
If all I care about is having people say they "like" my image, then my work is done as soon as people say positive things. But if my goal is to convince people that something fake is indeed physically real, then my work is NEVER done, so I keep striving on and on and on.
When two competing rendering engines are being compared technically there needs to be some unbiased metric by which to compare them. That "metric" will be the degree of technical realism, as the engine that can provide the most realistic image is assumed to be the engine that is the better quality. When parting with hard earned cash most people want the best bang for their buck so if the unbiased engine is ultimately higher in quality then it stands to reason that so many people would seek out the unbiased engines.
Part 2
You need to ask yourself a couple of questions before you go any further. There are three aspects you need to address
1. Quality
Is realism important to you in your renders? If no, then biased rendering will be just fine for you. If however you truly want realism, then unbiased is the only real option.
2. Speed
Very often biased rendering will "appear" to be faster than unbiased, but the reason biases exist in the first place is to speed rendering compared to unbiased methods, so this shouldn't be a surprise. Unbiased rendering performs many orders of magnitude more calculations per pixel than biased rendering, so it is almost never a good idea to try to draw any conclusions based on rendering time. Forget 3Delight speed comparisons with Octane, its like comparing apples to coconuts, completely unrelated species and highly misleading
To make matters ever more complicated....Unbiased render engines have the benefit of being modern, written with modern code and modern GPU hardware, so in many cases, unbiased rendering for all of its improved quality can still often be faster than biased CPU methods. For me, Octane is literally thousands of times faster than CPU. Renders that would have taken weeks for unbiased level quality are done in mere hours, enabling me to take more risks and use more clever settings and effects than I would have otherwise found too expensive such as depth of field, blurred reflections and the like..
And this is the real point. Unbiased rendering is much faster than biased rendering when physical realism is the goal. To take a biased engine like 3Delight and tweak it to unbiased levels of accuracy typically raises the render time far beyond that of a typical unbiased GPU accelerated render engine like Octane.
So if you want realism that is fast rendering, then Iray and Octane are the only feasible options. If however, subject matter means more to you than technical merit, then biased rendering is fine so long as you don't try to compete too directly with unbiased renders.
3. Ease of use and feedback
And lastly Don, it turns out once you get into it, that unbiased PBR render engines are actually mush EASIER to set up. You only need half the sliders to adjust materials. Realize that all that "expertise" we gained with biased engines was to compensate for the biases in the rendering engines. Once those biases are removed, materials definitions become so much simpler. Sure, it took YEARS to get good with 3delight, but that doesn't mean it till take YEARS to get good with Octane. Quite the contrary, you will be masterful within mere days to weeks. The learning curve is not "steep" by any means.
It's almost unfair how much easier certain things are in unbiased engines. I sometimes feel guilty remembering how hard I used to work.
I have read a few comments that unbiased isnt good for humans, but this is in my opinion extremely flawed thinking. Unbiased engines tend to highlight strongly any latent flaws in materials and models. Just because you used unbiased lighting doesn't mean the original artist didn't bake unnecessary specular highlights into the skin ruining your otherwise unbiased render. Sure, in 3delight having those highlights baked in looked awesome, but that same skin in Iray looks like a burn victim.
Fact is, most of the content in the store related to human skin is simply not high enough in quality to look good in unbiased engines. We need an entirely new class of skin textures, with NO baked in specular highlights (which is almost impossible since most real skins are based on photos and have specular already captured)
Also, after years of looking at thousands of renders, we've trained our eyes to have unrealistic expectations about a lot of things like SSS. Most people really do need to completely retrain their eyes for realistic levels, not exaggerated ones.
But this is only for those who seek realism. For more stylized works, there are no rules.
Rashad,
Man, these write ups are really getting into it, and you make some very strong points. That point about the skin and other materials being flawed is IMPORTANT. Thank you.
Let me go in a different direction for a moment.
I get the impression that unbiased renders create the picture IT wants to create for you… kind of a “what you get is what you get” sort of thing. Its so pristine and realistic, you hardly want to change anything about it. It is already perfect, so don’t you dare paint over it… you’re messing with reality here.
Whereas, again, I get the impression that a biased render gets you closer to building the image that YOU steer toward yourself, and in that sense, you create yourself. You’re eye is the judge, your sense of lighting, shadows, all that. In this sense biased renders are “more artistic” because you design the look of the final image yourself.
In you’re opinion would this be correct or incorrect thinking? Thanks in advance.
- Don
Its Art, realism does not come into it
want realism get a camera and some starving models oops :red:
some use toon render settings/plugins with line art generation so it is kind of a pointless arguement with no right answers really.
A lot of good advice here already.
It really is a matter of personal choice, and what works best for you.
Unbiased renderers have been around for a long long time, but until the past few years, have only been the realm of people who wanted the most realistic lighting possible (especially for caustics), and had a lot of time to render the image. In recent years (the last 8-10), improvements such as multiple core processors made traditional unbiased rendering more appealing, because all of the sudden you had more "processors" in your computer to use for rendering. Thus the popularity of LuxRender - a render engine that was "free", and was capable of outstanding realism and 3D light simulation
Then Nvidia introduced their CUDA technology (soon followed by Open Compute Language or OpenCL). With Cuda the average person could have hundreds of processors on their desktop using the stream processors on their graphics card, with little or no additional investment (basically giving them a mini super computer for calculation intensive processes). This was a huge thing in the scientific community where serialized number crunching that used to take super computers days to complete, could now be completed in hours ..... on a single desktop computer! For example, using CUDA I could run noise reduction filters (customized low pass filters) on huge digital elevation models generated from LiDAR data that normally would take 16-20 hours in less than 120 seconds using CUDA.
GPGPU computing is also very well suited to unbiased rendering, where the computations are (or can be made to be) very linear and can be easily optimized for efficient throughput for GPGPU number crunching. This meant that unbiased rendering, which used to either require many hours/days, or a small render farm, could now be done at a much faster rate, even approaching speeds that were being achieved with high-end/high-quality biased ray tracing render engines.
So what does this mean to the average user? It opens up the capability to use a render engine that uses a minimal amount of "cheats" to to mimic the real world properties of light, and it's interaction with the materials it is absorbed by and reflected off of. OK, so what does all this gobbledygook mean for the average user? It means that "in theory" it will be much easier to set up believable 3D environments with light(s) and materials that behave as they do in the real world, and therefor reducing the need for additional lights to fake bounced light (AKA Global Illumination). In my experience, the theory is actually very close to the reality of using unbiased render engines. However, using ray tracing and GI (Global Illumination) in Carrara is very much like using an unbiased renderer, but the render speeds of the Carrara internal renderer using ray tracing and GI will typically be much much slower than GPU based unbiased rendering, but usually faster than the curent version of LuxRender using CPU only or Hybrid CPU/GPU (LuxusCore CPU only will typically be faster than Carrara using GI/ray tracing at higher quality levels).
Another benefit of Octane, LuxusCore, and to some degree Iray, is the ability to have an IPR (Interactive Progressive Render) which dramatically improves the workflow when setting up lighting and materials (most of the big studios have either moved to GI/ray tracing and IPR, or are quickly moving in that direction - this type of workflow dramatically reduces the number of person hours used during production - but also increases the final render times). For me, what I really like about Octane is the IPR. It makes working on your materials and lighting SOOOOO much easier/faster. This is what I really like about LuxusCore too. The IPR in LuxusCore doesn't feel as responsive as Octane to me, but it is still very usable. The IPR combined with (for me) the simpler or more logical lighting and shader setup is what I like the most. An IPR is also available in DS now, with both 3Delight and Iray. The IPR with Iray isn't nearly as interactive as Octane, and it also feels like the IPR of LuxusCore is more responsive/faster as well. However if you have a high end machine (both GPU and CPU) the IPR with Iray will perform much better. I don't know what plans DAZ 3D has for Carrara, but I rather doubt that an IPR for Carrara's internal renderer is even on their radar screen. It only showed up in DS with 3Delight because it was available through 3Delight (i.e. relatively simple to integrate), and I would guess that it the same for Iray. My guess is that the only IPR's we will see in Carrara (at least in the near future) will be through the integration of external render engines.
An interesting side note here: Progressive rendering, ray tracing, and the use of IPR are actually becoming more of an "industry standard" (if there is such a thing) with the larger studios now. Many have already switched, while others are in the process. Though they aren't using unbiased rendering, they are using more GI and ray tracing which speeds up the workflow and requires much less time (lighting artist time) to make changes to lighting when changes are made to a scene (often required even with modest changes in camera angle when faking GI). But of course there is a down side - the final renders do take longer and/or more CPU time.
The bottom line it that everyone's needs and tastes are a bit different. After using Octane for Carrara, going back to using the internal render engine in Carrara isn't fun. I quickly remember how tiring the iterative process of tweak, test render, tweak, test render, repeat until you get the result you want (or you just don't care anymore!) is. I can't say I won't go back to the render engine in Carrara, because it is capable of things I can't do in Octane, but I can say is I prefer using Octane much much more than Carrara's internal renderer, and will only use it when I need to do something that can't be done in Octane (like Carrara dynamic hair).
OK, now to answer some of your specific questions:
• How do you classify render time when using unbiased (I know they can go on forever, what do you look for, a superficial “it looks good enough” or you getting in there with a magnifier or something)
When it looks good enough, you stop it. for animations (Octane), you set the number of.samples per pixel for the quality you want to achieve, when the number is reached, the next frame is rendered.
• How long does an acceptable quality 720p take (ballpark estimate, all related to animation rendering)
As others have noted - this is highly dependent on what is in the scene including materials used and lighting. Wendy or SiFi Funk may be able to give a better answer to this one, since they do a lot of animations.
• If I double the size of an unbiased render, does rendering take approximately x4 longer as it does with biased (double size is 4x the area)
This is basically true for any render engine. 2x more pixels will take approximately 2x longer, 4x more pixels will take approx. 4x longer.
• This is the big one: Animations and network rendering… possible for hobbyists… or not (I have good but certainly not high end GPUs)
Hopefully SiFi Funk will pop in and answer this one, I think he was working on doing this with the Ocatne plugin. Another option is to simply add another card (or 2, or 3) to your computer. I've seen posts of computers with 4 cards for use with Octane (here is a video of a system with 4 Titans :bug: https://www.youtube.com/watch?v=mhqf1n2xq80).
One last note. I actually felt the same as many others who have responded here when Octane first came out several years ago. Though some of the renders I saw were totally amazing, at the time I had Reality/Lux, and felt that if I really needed accurate lighting/cuastics, then I would simply use it. Plus I was very happy with the results I was getting from Carrara. Fast forward to a little over one and a half years ago, and there was an announcement that the Octane for Carrara plugin was under development. Out of curiosity, I thought I should check out Octane to see how good see it really was (thinking it was going to be interesting, but not something I would probably want to invest in). I downloaded the Poser plugin Demo (no Carrara plugin yet, and no DS plugin Demo). I was amazed that all of the sudden, in my opinion, Octane was actually more impressive than anything I had read/heard. The interactive workflow, along with the ease of use and the speed/quality was what got me hooked. Call it OOO...shinny syndrom if you want, but honestly, I hadn't had so much fun with 3D in years. I actually felt like I was in control of the creation process - while I was creating, instead of setting things up, rendering to see how things actually looked, then swiftly being swept into the "tweak, render, repeat" groundhog day syndrome for the next few hours. Every little change I made was almost immediately reflected in the ORVP (Octane Render View Port). I began to experiment in ways I had never done before, and was completely enjoying learning the materials and lighting. All because I was able to see how the final render would look with virtually every change I made, almost immediately. A month went by and my enthusiasm for using Octane was even greater than when I first started. My wife saw how much I was enjoying it, and told me to buy it. To this day, even with the free inclusion of Iray in DS, I don't regret anything I spent on Octane.
Maybe it's the fact that I'm not an artist, and not a Carrara internal render engine expert (I've always felt most comfortable with ray tracing and GI) that I find Octane such a great addition to my 3D tool kit (I also really like Luxus Core and Iray). But it really isn't a case of the new shinny toy for me. IMHO, the quality of my renders has improved, as has my interest in 3D since investing in Octane. But I'm just a mere hack (and have also been called a geek many times), not a true 3D artist, so take my "words of wisdom" with a grain of salt. But, with a quick look at my rendo gallery (or here to some extent), you might just come to the same conclusion that my renders have gotten better using Octane (or maybe not - like I said, I'm not artist, I just enjoy playing with this 3D stuff).
Sorry for the looooong post - time to go back to hiding in the shadows :roll:
More artistic is a hard sell for me. Indeed, the biased engine allows me to create more fantasy materials and such sure. But due to slow rendering and the need for lots of sample test renders I'm also going to be less likely to employ depth of field, or full GI, which to me are detrimental omissions for the artist. Any "option" that the artist has to rule out for any reason is to my mind a limitation being placed that shouldn't be there. So while rendering unbiased I might lose the ability to make things glow arbitrarily, I by contrast also gain the ability to render with DOF and other cool effects I would have otherwise avoided. So I would say my final vision is probably more realized in the unbiased engine than in the biased one.
And it is important to accept that most people new to CG "assume" on the most basic level their render engine is unbiased by nature... just like we assume the rights and wrongs we are raised with to be unbiased. Only when we get older do we look back and acknowledge the biases of our upbringings. To realize that mom said don't eat that because she doesnt like it, not because there is actually anything wrong with this particular food. They assume that if they add a light source to a scene that the light it creates will behave as it does in nature. It is only after spending some time with the tool that users then learn to know when these lights behave naturally and when they don't, and how to compensate for it.
Unbiased render engines are much more predictable in general, I agree. But that is the strength of it.
If you were starting out brand new today in 3d cg rendering with no previous experience with other engines, an unbiased render engine will be more intuitive and easier to learn by a long shot. Light behaves the same way in Octane as it tends to in real life, so as new user who only has experiences with the real world to draw upon, Octane gives results exactly like those you'd expect. This means that even with very little knowledge of 3d and cg, I can still churn out a decent quality render.
Again, this is all related to realism. If realism isnt the primary goal, then unbiased really doesn't offer any advantages.
With all due respect, that's just straight nonsense. But anyway...
Again, I think most of the responses that are posted in topics in this forum are from folks who are, in general, one man bands who produce stuff mainly for their own use or enjoyment. Not all, but most. And that tends to be a vastly different perspective from those who produce stuff for others. And I've said that before, maybe 10,000 times.
*IF* you need to produce something that others will see and enjoy, you need to have a purpose and goal for your image. And you need to understand what it takes to move them, impress them, place them in the environment you've created, make them feel the emotions you want them to feel, and be interested in the story you're telling. Which means you have to talk their language, and understand what that language is and what their expectations and needs are.
And if you want them to be scared to death and grip the arms on their chairs because there are some huge mechanical droids clomping down mainstreet past the 7-Eleven and Wal Mart, and firing lasers at a Ford Fiesta with a mom and her 3 year old kid inside, then you can bet you need to do a lot of work to make them believe, visually, that it's all real. Even though it's absolutely impossible, never will happen. And if, for one moment, they don't believe it, then you've lost them.
In that case, your art and the success of the film depends upon making the impossible seem real.
Does the viewer care one bit if you use an unbiased render engine? No. The viewer doesn't even know what render means. And if you can produce a sufficiently realistic looking image, like professionals have been doing for decades, even before unbiased renderers existed, then who cares? If you can do great compositing instead of using unbiased renderers, then why spend the time and effort to render for months?
On the other hand, if you want the viewer to spray his Coke thru his nose while he's watching your animations on South Park, does it matter what render engine you use? No, it's irrelevant.
Now, if you're doing an animated film like Despicable Me, do you need to have absolutely photorealistic rendering quality? No, because the viewer doesn't expect a photograph in an animated cartoon.
It all depends. On what you need, and what your viewer needs, and what you can get away with.
And like I say, if you're producing stuff only for yourself and your kids, then just do whatever is fun. And if you don't want to put in the effort to learn how to make a 90% realistic solution in a standard renderer, then there's no question. Try whatever renderer you want. Because at the end of the day, who cares what the final "realism" of your render is?
Honestly, I've seen people here produce renders that they are convinced look "real", and I just scratch my head wondering what they're thinking. So if no two people can agree what's "real" looking, and nobody wants to match their stuff to the closest thing to real we can find (ie, a photograph), then what's there to discuss? Do what you want, because anybody you ask will have a different idea of realism in the first place.
And by the way, before you believe that you NEED an unbiased renderer in order to obtain "realism", I suggest you stop, take a step back, and look at the last two decades of feature film production. Look at what has been produced by the CG/VFX industry in the past two decades using biased renderers, like Renderman, which is pretty much the standard renderer for many years in feature films.
Do you not think that all of those films were "realistic" enough? Did the audience not believe they were real enough?
Yes, unbiased renderers have advantages. And they have disadvantages. But they are not the only thing that will meet your goals if you're trying to produce something that is believeable.
Joe, from what I've seen you've never actually used an unbiased render engine yourself. How can you make assumptions with confidence if you are not yourself familiar first hand with the issues in discussion? Educated guesses are fine, but I want to hear your real experience.
If you were me, sitting on two titan blacks you'd be looking at unbiased rendering with a lot more excitement, just my assumption.