Rendering 101 - pt. 25: AOVs In Redshift
Added 2020-08-06 07:00:04 +0000 UTCAnother one of those dreaded topics, but actually a quite useful one: AOVs. Or Render passes. Or "those nasty buggers I always forget how to set up". In order to help you (and Mo) remember how to use Redshift's AOVs, we recorded this quick video.
Comments
You guys should cover the light map feature in Redshift!
Parker Coleman
2020-12-12 03:28:22 +0000 UTCHi, thanks again. Having some Nuke Comp background... of course I like AOVs from 3D :-) My arguments to use AOVs: - It's helpful if you need to integrate live action and CG and other elements and additional CG renders..., you can tweak specular, reflection, refraction individually, and with Cryptomatte also individually per Material. You are matching the 3D lighting as much as possible like a 3D lighter, but it happens in Comp. - 3D Rendering, even using GPUs, is still '"expensive". Having the the shading and geometry data available in Comp can be a huge time saver, as it's more likely you do not need to re-render something, if changes are necessary, by client request, or just because of mistakes, errors. This is more important with long renders, lots of frames, multiple shots, few time. - Depth of Field, Bloom, Flares, Glares, Lens Distortion can be added in Comp, saving time in 3D, being able to iterate more often. If you do not like those effects anymore, they are suddenly baked into your 3D render and you need to render again. This has nothing to do with AOVs necessarily... but it's the idea about keeping a scene modular and quickly changeable as possible. But yes, for Motion Design and if there is no dedicated Comp Artist, it might be not useful; A lot of time just a beauty pass works, actually. Sometimes you have all those AOVs and actually did not need them all the time ;-) One hint: If the comper asks for AOVs, the first thing to do is to rebuild the beauty with the individual lighting/shading passes setup in 3D, which is described here, for After Effects, Nuke and Fusion: https://docs.redshift3d.com/display/RSDOCS/Integrated+AOVs?product=houdini#IntegratedAOVs-CompositingStandardAOVs I always wanted to do a quick render comparison about render speed with and without AOVs: Rendering the above "start" scene from Frame 1-24 (first filecaching those frames in /obj/grid1/filecache1) with a Laptop nvidia RTX 2080 Max-Q, from and to an SSD; standard EXR settings (zip scanline), 16bit: Without AOVs: 59sec Average Frame Render Time; One EXR file with 7 MB With 6 AOVs (4x shading and 2x mattes): 88sec Average Frame Render Time, One EXR with 25MB; Cryptomatte with 13MB. With 14 AOVs activated (shading passes, and P, N, Z, V geometry passes, 1 puzzlematte, 1 cryptomatte): 99 sec Average Frame Render Time; One EXR from 30-49 MB plus extra EXR Cryptomatte with 13 MB So it's 1.6 times longer and more expensive with 14 AOVs, but the longer the 3d render time itself increases, the less it should matter: If I "invest" in a full quality production 3d render, having AOVs gets the most data out of the 3D step, to hopefully make any adjustments in Comp. At least I should get some mattes. But the faster the 3D setup renders depending on render power... I also like it to iterate in 3D and change stuff there. It, as usual, depends on the project, people, briefing, scene, time and budget.
Martin Backhaus
2020-09-30 02:31:37 +0000 UTCLove this tutorial. I am comping my renders in photoshop using EXR-IO to import the AOV's . Is there anyway to make sure the AOV's have an alpha mask? Any idea if I can export as an EXR at 16 or 8 bit as I can't do much in photoshop with these 16bit floating things and converting them compresses everything.
Thomas
2020-09-28 20:59:53 +0000 UTCI do a lot of NPR style rendering so I'm always saving out puzzle mattes and weird masks to deal with in comp. I'll definitely try the custom ID masks because that seems really powerful to me.
Thomas Roohan
2020-08-31 23:24:46 +0000 UTCA general overview of glows/blooms, LUTs, grain, defocus(depth), chromatic aberration, denoise using Houdini with or without RS would be great.
Joerg Fluegge
2020-08-21 05:13:29 +0000 UTCRegarding workflow, I always do either touch-ups in Davinci Resolve or full compositing in Fusion, using primarily puzzlemattes for object masks. The main reason for color grading is to fine-tune the superwhites (which are easily blown out) the same way we use color profiles in cameras. So yes, no more long lists of AOVs as we use 10-years ago, but still use them everyday.
Olivia Cefai
2020-08-17 17:44:22 +0000 UTCLayout and compositing approaches for large/complex scenes has always been something I want to understand from an industry standards perspective. I assume thats where it really shines. The amount of time it takes to cache and render some of these detailed sims strikes me as something you would want to break up for the sake of time/flexibility. Plus id imagine being able to juggle elements from other software's would be a major strength. Would definitely be interested in more "theoretical" breakdowns relating to these subjects in the future.
Elias
2020-08-11 00:29:04 +0000 UTC
