Image Sampler

[glorybound]

On an average, how much time does applying an Image Sampler add to your render times? By chance, to see why my render times were so high, I turned it off. Now, my render time is 1/2 what it was when it was enabled. Is it one of those things best done in post?

[joconnell]

Which sampler were you using? Something like catmull or mitchell both increase contrast between edge pixels which is kind of similar to the difference in grain that you're trying to remove using colour thresholds. Blurring filters in theory should make it easier to sample something as they're averaging values closer together.

It depends which one you're using really - I render sequences so something too sharp can be too strobey for me, and if I render anything too softened it might be blurrier than the footage I'm trying to match to. If you're doing stills where dancing grain isn't an issue, then I'd definitely do blurring or sharpening in post after for flexibility. You'll get a little speed benefit in most cases but half sounds bananas!

[super gnu]

are you talking about the image sampler or the aa filter? turning off the image sampler ( setting it to "fixed, 1 subdiv" ) will definitely speed things up but also make everything look dreadful.. this could be compensated by rendering at many times the final resolution and downsampling in post, but thats gonna be slower than just using the image sampler.

i assume however that you are talking about the AA filter. i remember many years ago trying to create the same effect as the "video" filter in post, as i hated the idea of rendering my stuff so blurry, but needed it blurry for the final comp (ah the days of dv footage, pal res and crt screens with pixels the size of jellybeans)

long story short i found it impossible to get the same effect, and moire effects that were reduced significantly by the video filter were exaggerated horribly by post blurring. i think its something to do with the aa filter working at the sub-pixel level.

as Joconnell says though, this was for video... for stills you have a bit more flexibility as to wether to use them. i would say watch out for moire patterns in your fences and stuff..

[joconnell]

Ah yes - the fun of dv - blurring was kind of hard on that with dv saving the colour info at half the res of the brightness so you'd have to split the image into luma and chroma, blur the two and then recombine. Thank god we have stereo 4k now!

[glorybound]

I usually use VraySincFilter, and I usually do architectural stills.

[joconnell]

Sinc is a sharpening filter so it'll make grain more pronounced, you might be better off with area or no filter and then unsharp mask in post.

[Neilg]

I never use a sharpening filter, they're slow and can cause issues. If you're rendering high res you don't really need one. Our work is either scaled down for web or printed at such a high dpi that any sub-pixel detail it picks up is irrelevant.

Been getting a lot of mileage from video recently with some camera tracked stuff. Just doesn't work the same if you do it post. But that's native res, and much harder to get away with not using one.

[glorybound]

very good to know, and doesn't seem to be worth the hit at render time. I naively thought, in some way, it speed up rendering by blurring pixels.

[Morne]

for stills I don't use any, but for animation I use quadratic