Page 1 of 1
Settings I've found efficient for MLT
Posted: Thu Jun 05, 2008 12:33 am
by CTZn
Hello, I thought I'd share these settings I've been happy with. They are a good equilibrium between a patchy and a noisy path, clears faster etc...
Tell me if that works for you, also be aware that they may be scale dependent (I'm not certain thought), IIC they work best for real scale.
Just set or edit or append:
Code: Select all
<renderer_settings>
[...]
<metropolis>true</metropolis>
<large_mutation_prob>0.05</large_mutation_prob>
<max_change>0.035</max_change>
[...]
</renderer_settings>
The light path will be very large and diffuse.
Posted: Thu Jun 05, 2008 1:40 am
by manitwo
works great. thanks CTZn!
Posted: Thu Jun 05, 2008 2:16 am
by Zom-B
afaik for SSS materials reducing max_change from default 0.01 to 0.005 is even better.
As far as I understand reducing the max_change parameter will keep the "rendering power" on the spot, once hit,
where a lot of work is to do ( => SSS Mat for example!).
Raising the max_change like you did CTZn will let the focus of renderpower jump more frequently in the image...
All Infos without guarantee of being correct!!!1!
Posted: Thu Jun 05, 2008 3:06 am
by StompinTom
yeah i guess raising max_change gives a more even distribution of the samples, meaning its perfect for the overall image and scenes with lots of diffuse materials.
Posted: Thu Jun 05, 2008 10:57 am
by CTZn
True, I seldomly have complex scenes out of the box, so these settings fit for diffuse preview mainly. And due to the lack of documentation/understanding this is highly empirical.
I'd use the term of focus like you ZomB, a lower max_change tends to a patchy convergence, that is higher sampling but also more localized (or as much samples but in a smaller area). Other areas are likely to wait more than with an higher mc for being evaluated.
Now it didn't come to me that different materials would work better with other settings, this is more than obvious now you said it StompinTom ! That's an interesting question... observing the correlation between diffusion parameters (in a general sense, and maybe as a mean of the whole scene), MLT parameters and convergence (?).
Posted: Thu Jun 05, 2008 1:08 pm
by StompinTom
yeah, for example, with lots of glass you would want the renderer to find a solid path and stick to it so you get clear caustics/reflections/refractions etc etc etc. but this might lead to patchy lighting/unequal sampling over the whole image.
so its a tradeoff, unfortunately, and i guess you just gotta have preferred settings depending on what kind of scene you got!
i think thats where Vray rules in terms of speed, as far as i know, it uses different rendering algorithms depending on the lighting situation within the same rendering.
Posted: Thu Jun 05, 2008 2:37 pm
by Wedge
Your going to make Ono hardcode these values...
Judging from his response now and then about changing default render values, it is not a good idea. It can cause 'render bugs' to appear for various reasons especially when people new to Indigo try changing them.
As far as I know he has never really approved having these available to adjust in an exporter (or at all) and as you can see the new versions of Indigo have a simple ini file. (read: much less settings than old ini files)
Just my brain dump from the past here. In my opinion if you want a quick render try path tracing or another renderer. I hope I didn't sound too harsh. (sorry if I did
) I think Ono just tries to avoid 'render bugs' coming into the forum that relate to changing values that should not be visible/known.
Posted: Thu Jun 05, 2008 2:58 pm
by CTZn
Even frustratingly short that was interesting...
Soz