Hello, I would like to know what are the setings for a flicker free animation in R15, I was much simpler in R14, cannot get my head around in the R15 verison any recomandation would be appreciate !! cheers !
Hi Sassi
The main problems seem to append when using Gi and the Physical renders, what would you use for animations with GI and Physical Render ?
These are my setting so far, These are the best I could have but using the standard renderer,
I would like to use the Physical render instead !
To my understanding the scene itself is highly connected to the settings. If there were a setting that would be “bullet proof” for ANY scene (with short render times), then MAXON would have included it.
That is why I have asked for details (scene-file) and your cache seems pretty small, but again—scene dependent.
Keep in mind that real light would take forever to render. The simulations in rendering are done with only a few “samples” compared to reality. You certainly know that, but I have to mention it.
Besides that, the question would be, is the flicker based on Physical or the GI, perhaps both? Your post reads as if only in conjunction the trouble kicks in. Is that correct—each by itself is fine? You wrote that you have luck with the standard render but you would like to use the standard render instead. (Physical Render instead?). I’m not clear if the Physical without GI would be fine. (I could only guess and that is not helping)
As I do not have the render power here (sorry about that), to just render many animation examples for individual scenes each time, perhaps you might check with the support.
My ideal would be, that there would be an update to the great series of Patrick’s Render tutorials for R15 (Physical Render). Perhaps you use the suggestion forum here.
Thanks again for you help Sassi, I will take a look at the Patrick’s Redner serie for sure,
Don’t get me wrong but as I remember the option indirect illumination in R14, was pretty neat in terms of rendering with a “kind of ” GI ? it use to works pretty well for aniamtion ! no ?
Anyway Rendering is a art by itself may we all become multi-task artist.
Yes Gregg, Patrick’s series is just great, clear and methodical. (Physical Render)
I try to avoid long render times and if I can get it set up with “classic” lighting, I’m happy to avoid it at all. Baking is certainly always a consideration. Everyone has his/her own challenges and preferences in work, it is a wide field.
I can’t totally agree that R14 was a push a single button solution and done. I am memorizing quite some questions that we got here. If the settings are to low (samples, rays, bounces etc) then there is always a danger that something starts to fall apart. There is a legacy option, does this work for your needs in conjunction with the Physical render?
I have done some rendering and I had the impression that R15 is faster and cleaner. Again case-depend. https://www.youtube.com/watch?v=RT8UEqKUaIE
Three clips, the first one with a RED camera, and the two that followed with R15. I think the physical render delivers here. Use the HD settings. There are minor problems based on YouTubes compression, if watched here in “raw”, they do not show up.
I agree as well that Rendering is a challenge in its own rights, if quality and render time should be in a nice balance.
Hey Dr. tahnks for showing me this
Yes I agree the Physical render is awesome in those shots, I think the main problem with GI and Physical render is when you have a lot of moving object
Hey Dr. tahnks for showing me this
Yes I agree the Physical render is awesome in those shots, I think the main problem with GI and Physical render is when you have a lot of moving object
The real thing that comes into play here is the limitations of the different GI algorithms out there.
First, GI is a blanket term.
It represents a way for us to render “bounced” light with “ease”.
From GI you have two main branches “brute force” or “caches”.
Now, “Brute Force” is much like it sounds.
It forces millions of samples (“rays”) out into the scene and collects brightness each time it bounces off a surface, this is then translated to pixels on the screen.
This happens for each frame of the animation.
Since every frame is a new calculation animation is not an issue, but the trade off is time.
Now, you then have caches.
A cache is actually very similar, with 2 key difference.
It can store samples for re-use AND it can average the stored samples together.
Now this is absolutely awesome for getting speedy renders but you lose accuracy in results.
Part of this is in the “averaging” process that happens to the samples.
This process is 2 part.
Part 1.
The scene is divided into records.
Records store a collection of samples, and display the average color of all those samples.
Part 2.
Records are blended together.
This blurs the hard edges together, but can also eliminate small detail.
Now, with a cache the scene tries to maintain a constant structure in the records that it creates.
As each time a record appears or disappears it can cause an “artifact”.
This happens often with object animation because the object will cross the borders created by the records, and force the record to rebuild.
Now, this doesn’t mean that this is totally unsuitable for object animation.
What it means is that you have to have record settings and sampling settings that are refined enough / high enough AND have record interpolation that is low enough to not totally mess up as records appear and disappear as the object moves over them.
Doing this increases rendertimes again, BUT generally they will be faster than QMC.
generally is the key though, in some cases, especially with small details QMC can be faster, or at least on par with the cache methods, BUT provides better results as grain is more desirable over splotches.
Either way, generally you are better off to try and get results that look correct (as Dr. Sassi is correct to always remind us). GI is a huge beast and needs time put into learn what each setting is doing for your render, and even then sometimes the only solution with GI is to wait longer (It is nearly impossible to speed up a scene that just needs a ton of samples). So in many cases, especially for freelance direct lighting is still preferred.
Also, GI is faster and it is used more in feature films, but this is because they are will to spend hours rendering a single frame. :D
Would you suggest to have an additional camera “observing” the problem area (as a detail camera) to increase for that area the cache information (always rendering to the picture viewer of course) to gain more cached data—not only for the Physical render—as well?
Which setting would you use, if, for the additional detail pass then?
Perhaps several additional cameras and skipped frames (?). So far I couldn’t find a “one size fits all” procedural.
Edit: this is all to gain more records where the flickering was before. Considering the settings are done close to the needs of the scene in the first place/edit
I ask, as I’m not certain if that is the solution for the problem that Gregg faces currently in R15.
All the best
Sassi
On the bottom plane is a trapezoid with a finer ‘detail” which was effect of the detail camera only. To illustrate the effect.
Thank you Gregg for the nice feedback, very much appreciated. :o)
I completely forgot to ask, is it flickering as in noise or more like in “splotchy”.
If I was under time pressure, and only noise was the issue, I used one of my noise filters/plug ins. I have several, as each has a different strength. QMC is normally noise for example. (I have invested mostly in them, as my practical footage needs it when filmed in low light etc.)
The “splotchy” part is based on to little “density/records/rays” of the GI calculation. I wouldn’t think about “fix it in post” at all.
Would you suggest to have an additional camera “observing” the problem area (as a detail camera) to increase for that area the cache information (always rendering to the picture viewer of course) to gain more cached data—not only for the Physical render—as well?
Which setting would you use, if, for the additional detail pass then?
Perhaps several additional cameras and skipped frames (?). So far I couldn’t find a “one size fits all” procedural.
Edit: this is all to gain more records where the flickering was before. Considering the settings are done close to the needs of the scene in the first place/edit
I ask, as I’m not certain if that is the solution for the problem that Gregg faces currently in R15.
All the best
Sassi
On the bottom plane is a trapezoid with a finer ‘detail” which was effect of the detail camera only. To illustrate the effect.
It’s a little difficult to provide exact details on settings.
But generally if you have a lot of animated objects in your scene, and a lot of fine detail then QMC is the best way to go, but it will be slow.
In this case Light Mapping or Radiosity Maps can be used to help with secondary bounces so speed up the results (Both of these can lead to splotches it the settings are too low.)
For cache based solutions the records need to be small enough that the viewer doesn’t notice when they pop.
The record smoothing value can help blend this out but then you lose detail.
So this one is more of a balancing act in finding settings that are still faster than QMC, but high enough that you maintain the quality of the lighting.
In the case of using IR, the secondary passes that work best are IR, Light Maps, or QMC.
(I have never tried using a “witness” camera for working with the Caches :/ so I can’t comment…I had always thought that the cache was only ever built from the Render Camera, which is part of the reason why animated objects are so prone to breaking the cache. I know in Vray you have the option of exporting caches and then merging caches together from different views, which would be very handy for many things.)
I agree with “It’s a little difficult to provide exact details on settings.”, Patrick.
The scenes I got so far here in the past years were mostly based on a single animated object. Yes, you are right this might not be the case here nor the standard (I admit ;o). In the moment I read your lines, I thought of all the “dynamic crowd spheres examples” that some people do. A detail or witness camera is not the solution here. I go just from the “render-settings” where the record amount is provided. But yes, QMC is more based on each frame.
I hope to get my hands on a new MacPro when it is out (the “can”) to do a more appropriate testing. Currently my resources are a little bit limited.
Gregg, perhaps not the case, but if you use objects as light, do you have switched on Material>Illumination>GI Area Light. Just a thought. (It is as in the image by default off) This can create a huge difference. A=on, B=off
As a side note: If I have to film (practical) in low conditions, or with strong highlights, the darks get noisy. I switch my RED-Epic then to HDRX to save the highlights, but expose to the dark areas, to get enough light. I get perhaps an over exposed image, as long nothing is clipped, I reduce in post the exposure about the 2 or 3 stops and my blacks are then clean, as I dialed in that way as well the noise level down. (I will publish soon a short where I used this a lot.)
This won’t work of course in GI, to crank up the light intensity to max and level it in post down (considering you work in 32bit/c for that). To illuminate the dark areas, where the light is not reaching to, needs more “bounces” of the light, or a Sky object to have a constant illumination all over the place, similar to an Ambient light. With enough light all over the place, dialing down the footage later might do the trick. Which means to avoid any “black holes” in your set up, as rays that get lost in that, instead of “collecting” some light information, deliver black. As everything is more or less “gambling” (hence the Monte Carlo in the name, for example) black hole information is then randomly exchanged with object illumination. Flickering is the result.
Portals and light objects to get the scene more stable is certainly helpful.
Here is a link with a scene file. Very simple, camera moves, cube moves and a tiny light source. Perhaps not totally worst case, but certain not an average scene. Lots of black around, and object and camera movement will easily show weak points in the settings. I had initially set up this file to explain the importance of the “GI Area Light” setting. See above. Easily to check, even in a single image. A/B.
More importantly, I have tried to get something on my laptop that is not bad but certainly shows that monochrome objects (banding) can illustrate the problem much easier. Errors and Problems hide better on structured textures, hehe.
Anyway, to see what happens with your settings just apply the settings you gave above to the scene and compare it with the movie file I have attached as well in the link. Again, I see some problems, but GI and Physical render, with no baking nor shortcuts, just a single light source was the target here. Stress test if you like.
I’m curious how others do it, I do not even pretend to have it perfect. ;o) BTW, I had set the “Lens” to an F/2.8, just standard, nothing extreme. I have deleted all my test cameras, only the main and the #4 (four.detail) is left.
Have fun.
Sassi
P.S. normally I would add some tiny bits of noise over the image to hide the banding, but for the sake of the “game” I don’t. Looks fine in higher bit depth though.
Hey Sassi today i’m finishing a crazy project, but for sure i will get a shot to testing these scene, thanks a lot , I will get back with result !
cheers
G.
I hope you enjoy the ride. GI is of course all about light. Which is normally my favorite theme as a Cinematographer/Photographer, and especially since I started early on to do many years light on stage for bands.
Light is the element that makes things work or break them in that area. I have seen a lot books about, but I have to admit, most people do not evolve over the three point light (which is only a start point, not a standard, a wide misunderstood) Well, I could go on about, but light is a theme that needs training and is not done easily, if one is after quality, which needs the inner understanding of many principles, not only shadow and density, etc.
The world of 3D is a different beast and if the real world light is just forced into it, it might fail. It needs a translation. If the limitations are known, the counteracting might be easier. It’s a long theme, one day I will do an appropriate series about. Perhaps too long for most people, I know, but professionalism has not really short cuts, right?
Just as a hint, in GI we stop light rays on their way, if we leave it that way, we might run into problems. As in my practical work with such, Light in a studio (which compares to C4D) might have some light sources that support, but did not even show up directly as such… hint hint. Have fun exploring it.
