Is this rendering technique technically valid?
fred9803
Posts: 1,559
in The Commons
I render a scene and it gets to say 70% convergence and most of the scene looks just fine, except for say a 10% area that is stlll grainy because it's in a low lit area that needs more time.
I stop the render and save the image, then I do a smaller spot render the problematic area, which then renders quite quickly, because I assume all my resources are now devoted to this small area and not the whole scene.
Then cut and paste this onto the original image.
It seems to work for me or am I just kidding myself about this time saving technique from a technical standpoint.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Yes it is fine. If the image quality is sufficient for your use, then that is what matters.
Technically, letting the render continue on the entire scene might bring about subtle nuances in the remainder of the scene, but again, if you are satisfied with the image, then it is fine.
Sometimes it is necessary. There is no way my system could do this image in one big render (not enough memory for one thing). In fact, it was split into at least 7 renders and then spliced together.
Classroom
If you are rendering in Iray just render the same thing at 3 to 4 times the resolution and scale the image down in Gimp or Photoshop. If this is anything else I suggest fixing anything you don't like with Gimp or Photoshop. The first trick only works on Iray. Iray does not care how large the scale of the render is so the scale down trick will give you great renders at half the time.
So does that suggest that if I did a very large scale render in Iray, it would take the same time to render as a smaller one? I'm a bit comfused about that.
A lower quality render at a larger scale perhaps. Is that what is being suggested?
Never worked for me - the larger scale always takes longer than the smaller. Whether that is pro-rata I've never bothered to work out.
My (admittedly limited) understanding of IRay is that it processes light information for every pixel, and my anecdotal experience is that higher resolution renders take longer to reach an equivalent level of completion. I believe the strategy of rendering at 2X size and downsizing in post uses lower render quality to reach completion faster (fewer iterations makes for shorter render times given the same image). The reason the strategy works is that issues like noise in the shadows and fireflies get averaged out when the image is scaled down. It's kind of like how going in the opposite direction, a small profile pic can look fine, but if you enlarge the image it becomes pixelated. In the first case you are sampling out small artifacts and in the second, you are in essence introducing them. Not exactly the same thing, but you get the idea.
I've done that a few times because I don't like leaving ANY computer alone to happily do what it's been told to... Mac or PC... The moment I get too far away for it to see me, it will do something stupid (I don't care if it's a $50,000,000 super computer)... "Could not complete requested action because baked bean level not authorized... Ok?".... What? WHY?.... Is there a critical level of baked beans required for a render? I don't know, but I don't want to find out, so if I have to go away, I'll fill in the fuzzy bits later.... Of course you probably can't use a render like that in a "No Postwork" contest, but I don't enter contests... If it were up to me, I'd allow it because in my opinion it's the same render, not an enhancement... But whatever.
all that really matters is are you happy with it. Does not matter how or what you did to get it the way you like. Many do several renders then cut and paste in their photo editors to get the image they want as well as post work with photoshop type brushes and adding photos in the background or other areas it's all about creating what you want to see using every tool you have at your disposal so long as for example if you share your image here for example it's mostly daz studio created. As for rendering length well most of the time I go 100% very rarely don't but even then there are times in some areas there will still be noise which you'll still have top do touch ups for in post work unless you like it or can't be bothered fixing or can't so your way is just as good
Yes, basically, if it works for you, then it's fine. Last time I checked the Render Police can't make you follow the text of the DS Manual explicity in the privacy of your own home. They just keep you from posting naughty pictures on the forums. :(
I do that a lot.. usually the skin around reflective jewelry takes quite a few more iterations to not look grainy.
It'll be much faster. I've even started to switch off the firefly filter. If I want a 1280 render, I'll render at 4096 and reduce in Photoship to 1280. Less than 10% convergence usually works fine.
Perhaps this is what I have not been getting about this method. Are you saying that you increase the size and reduce the convergence percentage? I've just been increasing the size without any further adjustments.
http://buerobewegt.com/quicktip-rendering-even-faster-in-iray/
I've read that article before and only now do I notice that it talks about stopping early (equivalent to reducing convergence). I assumed that it meant that the render times - at full convergence - be similar for larger sizes.
I would say that whatever gets you to your point B is valid, no matter what it is ;).
Laurie
Thanks everyone for your comments. At least with Iray I can now do this, and nor wait for every little firefly bucket to render like the old days.
Did i miss if anyone answered the actual question......Is Fred actually saving time by stopping the whole render and then just doing a partial render of the area and compositing later?
Yes, I do that all the time with renders where I use reflective accessories, it saves a lot of time. Like hours and hours in total on sets of promos.
I find the low to mid 80s is fine, with some shaded areas needing spot rendering; I combine them together afterwards.
@OP
The answer to your question is: are you happy with it?
That's right: increase the render size (don't be afraid to go to 4x larger than your desired output), and stop it early. I usually only need to go to 10% convergence or less. Once it's reduced in Photoshop, etc, the reduction filters out the fireflies. You should also try switching off the firefly filter to speed things up even further.