I’ve been trying to get some nice, crisp GI renders and been struggling in the removal of the “splotchiness” from the IR cache. This is occurring locally and on Net Render. I’ve tried a few different ways. One, just use IR maps, no prepass and crank the IR samples (helpful, but not 100%, even at 5120 samples). Two, Net Render animation mode, with IR+QMC (NET Render), a 5 frame skip for Pre-Pass, sample count set to “High” and Record Density set to Medium. But this still results in blotchy renders. Even if I take the the sample count to 2048, I still get splotchiness. Now, my question here is should the sample counts be up in the 10k range for sweet, buttery, smooth, silky renderings? Or am I missing something else completely?
Attached is an example file using a GSG preset. Of course, if I ZIP the working folder, I get this error: The file you are attempting to upload has invalid content for its MIME type. So attached is just the scene file….why does the attachment system here suck so badly??
I had a short (still—not animated) look into your file. The .hdr file is missing. (I have replaced it with a gradient, here an .hdr might be OK, as we talk about just a simple fluorescent. My suggestion, avoid this format at ALL cost. it is a 4 time 8bit format only and has a horrible color fidelity. Use Open EXR)
All scene parameters are on default, and the lights looks like copied in (only reflection is set on to 2%). The fourth wall was off, which introduced a a lot of “zero light” values in the scene. GI is based on randomness as it is impossible to cover the calculation with the same density as you find this in reality. To get renderings in a reasonable time, this randomness has some parameters, to get the scene specific needs covered.
I saw the gamma set to 1.8. Since a while even MAC is natively on 2.2. The time 1.8 was used, was based on a LUT (Look up table) prior to the scene that introduced the need to have anything else set up to 1.8, to have on the need Macs working with screen that had 2.2. As an result -during the change, QT had to be set up precisely for the making it was used on. 1.8 for renderings is not in use anymore. This scene has Linear workflow disabled, which I have to admit, is nothing that I’m used to anymore. Which means that any texture with an baked in gamma (1.8 or 2.2) will appear as gamma (not corrected) and the light energy given to the scene will be gamma based. On the end—for viewing purposes you have a gamma AGAIN applied on it. Twice then for the part that affects with its backed in gamma the scene.
If I switch the 4th wall to on, I have the impression it looks not bad or splotchy, which is certainly a subjective call on my side.
For animation, the AntiAlias shouldn’t be on still. (Yes, I assume that you know that for sure, but this is a forum and I think it is worth to mention) However, the idea to just crank up the light intensity and leave out anything else is certainly nice for the glossy look, but should be taken with care. Why? Because the Antialiasing options work not very well with super “hot” and strong contrasts. A reason why I use for my >=100% light sources a gradient. One might argue that a ligh source has a clean edge, but besides Hyper-realistic (CGish) approaches, most lenses will not have the resolving power for such “HDRIish” contrast, and “light glow/light wrap” has to be introduced. Which is not an option in C4D. So post production has to care about that. Rendering is just one step, it shouldn’t be the final. :o)
As the camera moves and no light nor object moves, to render this scene at all in full animated GI seems not really economic at all. Bake the light into the object and leave the reflections on.
I rendered it as given with 1.8 gamma and linear workflow off, dull and not splotchy (again that might be a question of taste and expectations)
Then I set the scene in the GI settings to 1.0 and the Linear Workflow to on and the scene looked really like a Office scene, clear and nice.
My suggestion, read up on the Linear workflow, it is the way rendering is done these days. Based on my subjective impression, you just started with some ready mades in GI. (Sorry if I’m wrong) and after some trial and errors, you like to have a fix here. GI in C4D is simple, but not push button simple, it needs some experience and knowledge. It needs certainly to know how to set up materials and light. Render-workflow etc.
Please have a look to the image below, and compare it with yours, as I did not had the complete scene.
Sassi,
Thanks for your feedback. I will try rebuilding a scene…as noted, I actually just grabbed this one from a GreyScaleGorilla preset and hadn’t modified anything but a camera. I did this because I was having the same problem in another scene (splotchiness during animation) that I built from scratch. I thought I’d just give it a go with another file that was simple and small. Let me run a few tests tonight on a new scene that I build and see what issues may (or may not arise). My immediate concern is that I am experiencing this problem across two files (both use GSG light kits) and I have had to turn the IR Samples WAY WAY up (5000+). Seems a bit too high, but maybe not?
p.s. HDR is missing because I can never seem to compress and upload anything on this site. I keep getting issues. If I compress and upload the scene file and try to upload the HDRI, then I get an error saying my post exceeds the allotted size, even though if I do the math on this end, it’s ok. I have no idea how to use this site.
My idea (in a nut shell) of GI is always driven by the randomness of the scene. Where can you help the “algorithm” to get with its few “rays” (compared to the real world) a more accurate and more importantly—a more stable idea of the scene. The first decision must be: is GI needed all the way through an animation. Not only to save render-time. Baked light has the wonderful potential to be easily adjusted (e.g., in Photoshop).
From global to local, as I would call it—The manual goes step by step through this decision process from wall to corners to detail. Besides that the materials have setting as well, it imperative to set the materials FIRST. There is very little on this planet that has 100 reflectance, and 100% emitting of light. I would go that far to say, there is not a single one. Then, COLOR should be not mixed up with Diffusion, one of the most not used function in C4D in the material editor. Some instruct even to use it just for dirt, which makes me always sad to hear. Anyway, if the materials have no reality, the scene will not have it either. Here comes a good knowledge of photography into the game, shut the automatic off and shoot. There is no better real world training for 3D than that.
The light source, yes a longer discussion, Light objects or Emitting sources. Well, Emitting source disable at all to use Phong, Blinn or Oran Nayar. Light objects have that potential. AS I mention in todays release (series) the Contrast in Light source is mostly not touched at all. Over a decade I got the impression people have never understood this slider to mimic larger r smaller light sources with this parameter.
All in all it is not just about sample amount. Sometimes you can support the application just by pointing to light sources with portals.
My suggestion, first material, then checking what light you like to have. From there, still of animation—and if animation, what can be baked, what can be excluded from the light situation. What render-intensive objects can be replaced (invisible) with a “dummy” to supply light reflected from other sources. etc. Special cases: What could be camera mapped? There is certainly more, but I hope that helps for the start.
I’m certain Patrick will chime in with some parameter discussions. :o)
Have fun exploring this option, it is a great way to render, but it takes some practice, and experience in that theme can’t be bought (from my point of view), it needs to be explored.
All the best
Sassi
P.S.: I saw your output file settings. For a monochrome scene like this one, I can’t stress enough to avoid 8bit/c formats. All the work and then saving out on that, even for tests, 8bit/c isn’t really enough for a good post production, e.g., color grading.
P.S.: here a short little test. The only change in the provided GI settings: Gamma to 1.0, Cache refinement (not certain if that is urgently needed here) and IR-Camera Animation.
As the lights produce very hot spots, I took it into AfterEffects, applied a little Motion blur on the rendering (Motion Vector pass double size and set to Preserve RGB).
The main part was to set a levels to output 0.5 to have finally a curve that is easier to adjust for the highlights. Then a levels with an out put to 2.0 to set it back.
This is perhaps not so important for the clipped white areas, but for the AA values, to calm them down a little bit. Post is needed for such a use. (I had set the Gradient for the Fluorescent image replacement to Black and white, with a little tradition. Multiplied with your 275% of the channel is certainly a challenge for AA.
Again this is just “one go” with standard setting, in an non color environment. All done of course in 32bit/c float. Some more tests might decrease render time, but I would bake it anyway, it is a camera only animation. Only reflections are moving.
If you scale it up, you might see some reflections issues from the lights, the “LightBank” object, as this appears even in non GI rendering and while the spheres set to full reflection. I checked as well the Fresnel and the Reflection Threshold. In the moment that object is out of the equation, the dance of the brightness values stops. So, as we talk here about GI, I don’t care about such light boxes. I love to set up ALL my objects on my own and know what I have. A careful check how you set up your scene is always critical! I think it is the little light seam around the boxes from the ceiling -but again, your question was about GI, so to eliminate anything else is key to get an answer here. (All of that might be not visible in your scene, as I have no idea about the Fluorescent texture, which might cover the contrast with different brightness values.) Compare the results, the LightBank is dissabled for reflections here.
Sassi,
This problem I’m having came from another scene (and unfortunately, one I cannot share at the moment). But, in that scene I am using the GreyscaleGorilla Light Kit Pro. Normally I just build my own lights, but for whatever reason, I didn’t in this scenario. I wonder if they are part of the problem? I’ve attached 3 files that I cranked out tonight with different GI settings and sample rates all below 256. I’m not reproducing the same issues in this scene….
There is nothing wrong with the lights so far I can tell, my concern was more about the texture that I had put in. As I mentioned, I’m used to create always from scratch but that is just me. I do not critique any 3rd party products here, nor try to give support for them. The lights that I have seen in your scene are only collections of native objects and options in C4D.
If you have only a camera animation, what speaks against a baking. To talk about a scene that none but you can see is rather difficult.
What I can do here, is to suggest a problem eliminating process. Which is sometimes not as simply as it looks like. Swap out any high polygon secondary objects. Again, I don’t know your scene. If there is no change, change the light source. Use a Light object, then a polygon based light.
Check in the materials, if every material really need to generate GI. Most object have not really the surface quality and size to really do anything to the scene. To jsut leave everything interacting with each other increases the possibility of problems.
I assume that you went one time through all parameters in the render-settings and that you are aware what they do. I assume as well that you observe the pre-pass renderings and watch them while they render, to see how the data is presented. The prepass representation might give you a clue.
To talk about sample rates ... without a scene. Would you really trust anyone who answer based on that?
You are in IR-Camera Animation—right?
So, what can I do else, you show me scenes that aren’t resemble the problem, and you give me some numbers. I have no further idea than to suggest the support. There you have your scene in an non public environment.
I’m not open for private project support.
What is different to the secret scene is also not to my knowledge. I can’t really add to anything I have shared today then, I hope you see that I care but what else can I do.
Sassi,
I appreciate all your feedback. Unfortunately, some scenes I have to work on aren’t for public viewing, so I try to recreate problems where possible in new scenes. Sometimes just having feedback to suggest looking at things I may have overlooked will solve the problem, which is why I posted here. And yes, you and I have talked “private” support not being an option, which I agree wholeheartedly with you about.
I had a long discussion with Patrick about this (and he peeked at the original scene) and we think we have figured it out. The problem may have been coming from the Greyscalegorilla Light Kit that uses an HDRI and “Use as GI Area Light” was not checked. I need to run a few other tests.
In my original scene, I had spotlights moving about. Perhaps if I had been smarter about it, I could have baked out the scene, then rendered matte objects that catch the spotlights. I didn’t think of this earlier, and probably should have. In theory, even if it wasn’t the best option, rendering out the scene with full GI should have worked no problem. This was what I was stuck on…
As always, I appreciate your responsiveness on the forum. You are a great asset and give fantastic information!!
p.s. The reason I posted the other files that “worked” was more to illustrate the point that multiple methods of animation work, and something was clearly overlooked in my problem scene.
for the update. I like the forum as the information is useable for everyone accessible even if none is currently awake to answer ;o) So, thanks for pointing to the problem, crossing the fingers that you find the final answer then.
Currently I’m in research mode for some other series, but I think there is certainly a need for a GI workflow series (this requires, from my point of view, a more tool character than an passive information base). Which I think will not work as manual visualization, or a flat parameter discussion. It needs to be more structured. My idea is more like an interactive decision and analytic tree. My impression after we have a new GI system since a few years, it is not done with the knowledge of few parts, the “interactivity” as well the flow of decisions based on characteristic analytical steps is needed. In other words—a problem solving strategy (or set up and understand the scene strategy). I will think later this year about. I’m not certain if linear based media are useful here, e.g. movies. Text as in the manual obviously doesn’t work very well anymore with most people. I think movie based tutorials have introduced some kind of impatience to the word/picture based only information. I believe highly that the next step is interactive, but that puts a lot of more effort in the creation. Which on the other hand is rather expensive, considering the periodically updates. As usual, I try to find out what is needed for “tomorrows” educational system.
Indeed. Knowing what to use and when is probably overlooked. Honestly, the idea of baking the GI didn’t even occur to me, because most of the time it’s irrelevant for what I am doing. Maybe some general “rules” or “guidelines” could be addressed. If XXXXX then YYYYYY. For example, when you would want to bake GI and when you wouldn’t. Or, when you want to do a pre-pass, and when you wouldn’t. Granted, experimenting on small scenes that conceptual addresses the larger, more complex scenes is a good idea. But in reality, I don’t always have the time for that.
Time is certainly always an issue. I understand that. As an very wise man told me in 2002: where you attention goes, your energy flows. Learning or experimenting is a vital part in this industry. To say it plainly, the less someone knows, the more time might be wasted. (...which could be turned into experience—to think positively) Of course, to know too much might cause to a lot of concerns, but it is certainly better to choose from options, than to get desperate with no option at all.
So what is the way to go then? As I mentioned it more often, one has to go through the complexity, to own the option of simplicity. To just take few points and call it simplification is just “underexposed”.
There comes exactly my idea of learning into the game: to supply more background than needed, while learning, and show connections among the parts. From there, a healthy base can be established while applying the data. On the end a massive and stable experience is created.
This causes a lot of research, like I do it at the moment with a texture series, “snap a photo and clean it up, use it as texture”. Well: fail(!)—but how to do it right and sustain quality is a longer story. (COMPLEXITY) To compress the needed time, instructors do the research and put this into tutorials. Make aware of the problems and allow the artist to find the best way after the training. I do not expect that anyone reads many thousands pages of color science and advanced photography books, but there is a need to know at least the results of such themes, if one likes to sustain quality. I know, many people like to skip that part, buy stuff and stay dependent. No my ideal of course.
Everyone needs of course to make a decision where to use time, and what the target is on the end. I decided to dig in everything I do as deep as possible. On the end, we are all responsible for the products we deliver. Clients are only happy until the deadline is reached, after that the understanding of time pressure ends, and the quality critic starts (most of the time) with out any excuses of “no time”.
Experience is the key to save time. Experience cost time. Experienced artists stay in business, un-experienced users are easily replaceable. Just my two cents to the “No Time” thingy.
And yes, you and I have talked “private” support not being an option, which I agree wholeheartedly with you about.