[Bf-committers] Cycles: Optimizing render speed?

Sergey Sharybin sergey.vfx at gmail.com
Tue Nov 20 09:10:46 CET 2018


Hi,

It is hard to predict which areas will be noisy at which frames when you're
working on an animation shot. With the manual control one would either need
to re-render the entire shot after giving clues to Cycles, or one would
need to store all the passes with all the extra information, and have some
automagic merged in place which isn't a great scenario for a small studios.
Also, keep in mind, with the denoiser in place Cycles already have a clue
which areas are problematic.

It is "just" matter of implementing good adaptive sampling criteria which
will together with the denoiser.

On Mon, Nov 19, 2018 at 9:08 PM Benjamin Humpherys <
benjamin.humpherys at gmail.com> wrote:

> What about a way to manually specify parts of the image that need more
> samples? Maybe there could be a “paintbrush” to paint over the areas that
> are a little noisier and need more samples. Or a way (if this isn’t already
> in Cycles) to specify particular objects or geometry that should receive
> extra samples. One edge case I remember being talked about in the AMA
> regarding adaptive sampling based on local contrast was objects with
> caustics. So in that case perhaps create a cylinder enclosing the object
> and where the caustics are expected to appear, and tell Cycles to take more
> samples within that volume.
>
> Basically, maybe the solution isn’t an algorithm that covers all cases,
> but an algorithm that works well enough 80% of the time and manual controls
> to fix the edge cases and provide artistic control.
>
> > On Nov 19, 2018, at 11:00 AM, Brecht Van Lommel <
> brechtvanlommel at gmail.com> wrote:
> >
> > We would like to add adaptive sampling, it's just a matter of someone
> > writing a good implementation which is not trivial.
> >
> >> On Thu, Nov 15, 2018 at 7:15 AM Kurt Weiss <fundev at kwnet.at> wrote:
> >>
> >> Sampling, Path Tracing:
> >> Samples: e.g. 70.
> >>
> >> Cycles renders all 70 Samples, regardless of whether the tile is
> >> changing or not.
> >> Often that's not neccessary in the main parts of the picture...
> >>
> >> Maybe it's a idea to add a possibility to define a threshold by
> >> comparing new an old images of the tile. (e.g. max. change of RGB value
> >> in pixels)
> >>
> >> Especially in areas where only background is present, the engine could
> >> abort sampling after one or two samples. This would save a lot of time
> >> at unattended rendering.
> >> _______________________________________________
> >> Bf-committers mailing list
> >> Bf-committers at blender.org
> >> https://lists.blender.org/mailman/listinfo/bf-committers
> >>
> > _______________________________________________
> > Bf-committers mailing list
> > Bf-committers at blender.org
> > https://lists.blender.org/mailman/listinfo/bf-committers
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> https://lists.blender.org/mailman/listinfo/bf-committers
>


-- 
With best regards, Sergey Sharybin


More information about the Bf-committers mailing list