I am experiencing a bad flicker in the animation of a scene I have created. They are appearing in the shadowed areas and in some of the reflective areas. As I increase the sample count the flicker seems to get worse. I have attached a link to view the animation and scene file. We have a render farm, so I am less concerned about render times and more concerned about loosing the flicker. Any suggestions on eliminating the flicker?
Sampling: Fixed
Sampling Sub: 7
HDR Threshold: 8
Blur Sub Max: 5
Shad Sub Max: 3
Ambient Occ Sub Max: 3
Subsurface Scattering Sub Max: 4
GI:
General: Primary Irradiance Cache; Light Mapping; Max Depth 30; Gamma 1; Sample Count 1024
IC: Record Density: Min -4; Max -2; Density 10%; Min Spacing 10%; Max Spacing 30%; Smoothing 75%; Color 75%
Light Mapping: Path Count: 15000; Sample Size: 0.005; Scale: Screen; Direct Light On;
Cache Files: Full Animation Mode On
Ambient Occlusion:
Samples: 1024;
Record Density: Min -4; Max -2; Density 20%; Min Spacing 20%; Max Spacing 50%; Smoothing 75%;
Full Animation Mode On
Difficult to say, the YouTube (even at 720p) has a lot of artifacts, or is that as well the flicker? Difficult as well without scene file (reduced to the minimum that will reproduce the problem)
Just some random thoughts:
You have three main areas combined, each can produce such problem, even I believe the AO might not the prime problem here, but that is an un-proffesional (as in guessing) assumption.
I would first shut off the Ambient Occlusion and the Physical Render, then render again with something better than “IC: Record Density: Min -4; Max -2”, it feels pretty low, compared to everything else. I’m not certain what you have outside of the 2 seconds, perhaps the Full Animation mode is not needed. Furthermore: Is every material that doesn’t produce relevant GI information excluded to contribute to the overall result? I have the feeling that some light is hitting tangentially the surface and that is why there is so little precision, of course - this is just one more guess…
The result will give you an idea if the GI is a problem or the two other parts.
The Physical is set to Fixed, any reasons why? Adaptive would be perhaps more efficient. Anyway, guessing about details, not my taste. As I can’t see with a movie what context is there at all. The materials are not known, so the long list of parameters is just half of the equation, there is so much more that can influence a scene.
The idea is to test each of the three (GI, Physical, AO) separately and where it might came from—GI seems the main candidate normally, so optimize here if the analysis provides this as result. Is there anything that can be baked? AO is certainly nothing that needs to be calculated for each frame, if at all. (Note, I’m a firm believer that AO and GI doubles the effect) I’m always a little bit baffled when relative reflective parts are included (if not for caustic reasons) Reflective surfaces live from reflection (mostly) - so baking would help.
Anyway, I might be completely off here, and guessing is something I really dislike, If you have a reduced file – I will take a look into it.
I see the YouTube link is gone and replaced with something that gives me SIX! PDFs that I have to read and to agree to. Certainly not, I’m not someone who just clicks on the agree button just because…
Perhaps you have a dropbox option, but a reduced file should be below 4MB (zipped) will work here.
Please have a look into this little file, it shows clearly the problem of light hitting (nearly) tangentially the surface.
You might adjust the red and the green light so it is more equal on the surface, but the effect might be the same.
Compare the “silence” of the green light and the nervousness of the red light on the gray surface.
If that assumption is correct, then excluding the gray, glossy material from receiving GI light might stabilize the rendering. A very low area light might help to deliver a tiny bit of light color bleed to those areas.
Thanks for your response. Not sure why you got SIX PDF’s to look at. That seems strange. The link was to a wetransfer zip file that is 15MB. Here is a Dropbox link to give you access to the scene file. I will also take a look at the other responses as well. Thanks for your input.
I made it through the file, but not a solution so far. I will attach a text which I wrote during testing. I might change my mind later, it is not a reduced scene… I’m sometimes tempted to suggest the MAXON “project-development support” of a “1on1” option instead, as this is way above a little forums help.
There is something in the model where I can’t put my finger on, even the large splotchy areas are way more silent, the set up of the model seems not clear to me. It feels sometimes like a little bit of a patchwork, done from other modelers and combined without adjustment. Am I completely off here?
I will delete the text soon as I think it goes too far into detail and I fear someone could take it personal… not my intention, checking the file since several hours.
My current best idea is to use not at all the full animation (“hall low” camera) and Radiosity Map instead of the Light map.
One thing that sticks out is the green light flickering in the reflections, which lead me to an idea (which I test currently): what would happen if I place close to the camera a sphere and bake the illumination that the sphere receives. This sphere would then act as a light source and would allow to lower the green light intensity. Similar to the white light from some small areas. The effect that I expect is based on the idea, that the small lights will produce a larger spot on the sphere and help to stabilize the situation. It is that smal patchwork of light that is seen by the GI or not, which is my current believe creates the problem, even with high density values to fight it. The sphere might harmonize that. Its like using china balls on set, you never see them and they just support. Similar to that could work a simplified background—the one behind the camera—not the scene background.
All the best
Sassi
Image: shows the Chinese Lantern cheat, which is invisible for camera, reflection and refraction during render-time. The texture baked for this is in the lower corner. Movie is attached in a later post.
Thanks so much for your feedback. The model was not one we produced in house and so your assumptions could be correct. I recently found where I was seeing a flicker in another scene, that two geometries were intersecting exactly at the same point. When I separated them, the flicker went away. Not sure if this is the same issue, but further testing will determine. I appreciate your feedback, and don’t worry, no one will take anything personal. The solution is always most important.
One render just finished, allow for a short moment to get this up.
There is one thing that I never have used, and which I think is used there very often, hence I have to explore it in a smaller set up, to check the point. It is not discussed in any tutorial here either, so it is time to clear that. Allow here for a moment as well. (It’s early here ;o)
I hope I sounded not like a complainer or critic about the model. If it would work, I would care about the set up at all. I know how busy production can be. But it took me while to see at least half of the organization in it. Hence the problems to see where it started to get problematic. The materials “talk” to me, and I see a lot of things have been tested and then checked off. But as I wrote in my text, most materials will not deliver at all relevant light reflection, so why bother to use them. Another point was, that pretty much every parameter was left on default. All of that doesn’t matter if the scene runs smoothly through the render engine.
Paul, please have a look to my test with the extra China-ball light idea and to separate the “not needed” parts with an larger gray sphere. I couldn’t resist to color correct and set the window with a quick blur back to “privacy”.
So, first take is raw, send is color corrected, etc.
I hope it comes closer to your targets.
About the other thing that I like to explore, after breakfast. ;o)
As I wrote above, I’m not used to do things like the “texture” version of this
example, and barely used the “selection” method.
So the scene is based on three set ups, a single little polygon is the only light source, anything else is NOT included as source (aka generate GI). The second one is already a little bit more sloppy, the GI light is not limited to the bright spot (selection alone. But the results are comparable with the first.
Where it gets really worse is the texture based “light source” with may dark areas in it. This might be not only true for gradient based things. The algorithm knows that is a light source and points more rays to explore it, but only a fraction is really useful.
All results are un-usable - a stress-test was the key here, and to see how sloppy scene set up drives quality down and render-times up. YUP, quality down and more time to wait. Sounds not like a good plan. To solve that with heavier GI settings sounds not at all worth the time nor effort.
Again, this is done to make a point, tweak it if you like for more data. Each scene follows more a typical scene set up, from clean to OK to sloppy.
The results are clear, enable only one for render and scene visibility in the Layer Manager (one of the top three entries), leave the GI settings for the first run alone, so they are not the influence. They are on default roughly. The white square for the initial light is similar in all cases.
So, my render-times: For a 31 frame animation!
Solo 03:21
Selection 04:16
Texture 11:42
And for nearly 12 minutes you get the worse result I can think off. See image on the left.
This is not fixable with more render power. A clean scene set up is un-beatable.
If you take the example “tangential” and the scene here, you have already most of the answers for your scene.
Further more, if you work with my “Chinese Lantern” cheat (hehe), then you should get good results. The cheat could be also just a large plane behind the camera, to simplify the camera background or other complex situations. The main idea is always, if you have a clear and simple (shader free, e.g. no fresnel in light source i.e. GI surfaces!) then things become much easier to solve for C4D.
My best wishes for the project!
Sassi Images, the left one shows the texture based results during rendering, again this could be a Fresnel or other changing influences, the right images shows the three single groups for the test rendering
Paul, please have a look to my test with the extra China-ball light idea and to separate the “not needed” parts with an larger gray sphere. I couldn’t resist to color correct and set the window with a quick blur back to “privacy”.
So, first take is raw, send is color corrected, etc.
I hope it comes closer to your targets.
About the other thing that I like to explore, after breakfast. ;o)
All the best
Sassi
That looks great! Thanks for looking into this. Could you post back the scene file with your “China Ball” approach. I am having trouble visualizing it and would like to learn how you achieved it inside my scene so I can apply it to other scenes as well. Thanks again! (And the color correct is actually close to our end look. Nice!)
The color correction was more a simple “Bleach Bypass”, as it felt right for this. Nice that I came close to your idea about it. When I expose with my RED Dragon, I go sometimes a stop over, just to put it down later one stop and kill eventual noise. With the China Ball I do the same, I give a little bit more light to work with, and then pull it down a bit. It is a cheat, but who will prove the artist wrong if the result fits.
The scene is here [see link below], with the baked texture. Note that there is a second sphere, which you might see in the reflections on the right. It is too small, but it was a test. Limiting the space for GI was crucial in the early days of GI in C4D (was it a decade ago?), so it is an old habit of mine to limit the studio space, hehe.
Reconnect the “Illum” folder!
The scene needs some work, but I can’t start over. I think I gave all needed information here. Besides that, Patrick has certainly shared a lot of tips in his tutorials. Have a loo!
Let me know if I should delete the download, as I imagine that your scene should not be too much in public.
I hope this thread was not too chaotic, but the scene was certainly a library of things to discuss. Thanks for your patience.
Thanks you for your time. It helps a lot. I have a membership on Cineversity, so if there are any tutorials you think would help in my understanding of baking textures and understanding any of what you discussed on a deeper level, I am happy to watch them. I like to learn as much as I can. Feel free to remove the scene links, but please leave the thread up for future “me’s” who find themselves stuck in the world of flickering animation. Finding these answers are like gold.
To remove the links was more meant to have your production not pre-published. I guess it goes into something larger.
Global Illumination is a constantly evolving theme. As you asked about R15, the series from Patrick will fit like a glove for you. http://www.cineversity.com/vidplaytut/global_illumination_part_01
He certainly presents the “must know” aspects of the C4D engine very well.
What I did here in this thread is to discuss a little bit what I would think is more often than not production based results, many people, different contributors and all more or less on deadline. I’m certainly aware of such scenarios. But if a scene has so many little render-time “black-holes”, the “render-review-redesign” cycle becomes long, even with a strong render farm. With that the realization amount what makes a scene work is lowered.
My suggestions should just lead to the question, is that this or that part needed in GI, or has it any real effect or value at all. More often than not I get the impression, that a good skill level in setting up lights are not valued, ... because we have GI and AO.
I know, I’m picky as I did in my early days many years stage light show (until I was 17, then I sold my truck load of equipment). I love light and as a cinematographer I love to set light in 3D as on a real set since long. It is a pleasure to do.
To bake things, to make my “long speech” above useful, is a step that can only be done—if the light works nicely already. It becomes less flexible, but way faster in rendering.
A tutorial that gives an idea about the power of baking is: http://youtu.be/Ows9cjiScmk
There are other ways to improve things, such as camera projection. Especially in the Hall_Low camera move, you could just take one image and re-project it for the material. The reflections and their change could be independent from this. It would also allow to paint on that single frame and have the light each frame changed. I need a little bit time before I have two more series up on my YouTube site. We decided to place them there to get a platform established between photographers and 3D artists. When all nearly 200 tutorials are up and tested enough, I will make a bigger announcement about, so far I keep a relative low profile.
I have made as well a series about typical mistakes (or creative approaches?) about light. I’m certainly old school, but a good idea about light is premium all over the place. ...if you find someone telling you “three point light” run! Thehehe. Sorry I couldn’t resist, but those tutorials are too often and people take them as rule, which results in uniform and meaningless results. http://www.cineversity.com/vidplaytut/a_short_discussion_about_light_part_01
This series was thought about a “download” of all my being picky—so I can start to work on a longer light series in a positive way. Perhaps I might start that series one day, so far no one had interest to support it. Yes, the GI option might stand in the way, as the solver for all light based aesthetically problems. We will see how 2015 will work for me.
So, I should stop here, light is certainly besides cameras and lenses a theme where I get “wordy”, thehe.
Thank you for all the helpful resources. After much, much more testing, I found that the ultimate solution came with changing the secondary method for light bounce to QMC as oppose to LM. It almost completely removed all the flickering issues. With more tweaking it gave us an excellent “non-flickering” render for that scene. Slightly longer render times, but not unacceptable for our farm. Thanks again for your time and helping point us in the right direction.
-Paul
SOLVED
UPDATE:
I just wanted to post this for others to understand. In order to achieve a render without “flicker” these are the settings I used. Keep in mind I have access to a render farm, so these settings are very time intensive for rendering but work well. If you are using one machine, expect long, long render times. I suggest starting low on the sample count and working your way up to an acceptable look. You can always turn up the sample count to get better renders, but with QMC, 128 seems to suffice for most things and may be overkill for others. These settings worked with lots of reflections and GI w physical render. If you are reading this and are new to C4D, there are other much faster settings that are better for still images or less complicated animation.
