Trying to figure ideal color depth flow chart for YT video production

[Walter Soyka]

If you’re happy at 8bpc, work at 8bpc. Not every project needs more.

That said, working at 16 bpc instead of 8bpc, even for 8bpc delivery, can be useful for preserving the appearance of subtle gradients and for reducing banding and rounding errors over multiple effects/blends during compositing.

If, for example, you are using a Levels or Curves effect and you’re seeing big chunky areas of similar color instead of smooth transitions, try promoting to 16bpc.

Working at 32bpc can change the look entirely, because it allows intermediate calculations to return results that brighter than white and darker than black.

If you’re stacking effects, and one of them blows out the highlights, and then another brings them back down, they will be flatly gray in 8/16 bpc because the first operation clips at full white; however, if you promote to 32bpc, the detail will be preserved because the mathematic values for the pixels are not restricted in range.

Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events

[Greg Sage]

Well, I think the 32 bit thing explains another issue I’ve been having where output was all blown out, but it looked great within AE. Looks like project was set to 32 bit, then when it output to settings I looked up for DVD, it was WAY over-saturated, so it must not have stepped-down properly.

I not only have TONS of gradients, but a bunch of color shifting where elements slowly rotate through the color wheel. It looks awful in 8 bit as it’s all jumpy and looses all the smoothness.

If it will help final YT viewing experience, I’ll be glad to work at 16 bit throughout. Not entirely clear from what I’ve read if YT is 8 bit only, or if the compression affects it in a somewhat less straight-forward fashion.

I do note that the Media Encoder settings they supposedly worked with YT to create says H.264 1920×1080 29.97 @8Mbps… but it’s been a bit harder to suss out what exactly happens to it from that point forward, or what exactly those figures mean in terms of the color bit depth.

Bottom line: If it the final output looks better for gradients and color shifts working at a 16bpc depth throughout the project, then that’s what I need to do. Just asking because there are so many layers that anything that can speed up processing is also welcomed.

[Chris Wright]

actually, there are a few more things. the nice 4:4:4 output dithered from After effects that makes all the 16bpc effects back into a 8 bit render, not so fast!

Guess what, if you use any 8bit effects in the 16bpc comp, it will break the dithering quality. and to make matters worse, if you upload to another site that requires recompression, your 4:4:4 will turn into 4:2:0. You won’t even know your video blew up till after you uploaded it!

you can first find out by test rendering out to a h.264 in adobe media encoder. that will show the chroma subsampling and 8 bit dithering.

if some shows up try:

1. fast blur/reduce saturation with broadcast colors effect as matte to master AE comp or rendered h.264 footage, then upload
2. render out both a 16bpc 4:4:4 and the highest quality possible h.264 and import, set as difference matte. use that as an alpha matte to blur/reduce sat.
3. try my AE project. you can test it out with test precomp working in 8bpc. it will only blur the quantization errors for 8 bit effects and low chroma quality 4:2:0.

ae cs5.5 aep
7137_168bpcwithquantizedchromabluraecs5.5.aep.zip

[Greg Sage]

A bit past my current comprehension level, but that’s good. Clearly I need to learn a few key items here. Just read the above half a dozen times to sort it out. Will dig in, but let me see if I understand the following:

So… 4:4:4 is synonymous with 32 bit color? … and 4:2:0 is 8 bit? So it’s 8 bit coming off the GH2? Output specs show 4:2:0. I ask because I’ve seen some conflicting info out there including reference to a 16 bit 4:2:0 output.

So, really, it sounds like I want to stick with 16 bit until final output and watch out for a few things, right? Speaking of which, is there a way to look at an effect and know right off the bat if it’s processing in 8 or 16 bits?

C4d has an 8 bit dither checkbox for output. This should be off if rendering out in 16 bit?

Starting to see now why my first dvd experiment looked like a cartoon gone horribly wrong on the burned disc.

Good stuff, and I’ll read up, but to clarify… when you say:

“you can first find out by test rendering out to a h.264 in adobe media encoder. that will show the chroma subsampling and 8 bit dithering.”

How exactly? What am I looking for?

For simplicity’s sake, btw… I’m only concerned about how the vid looks on my site, and on Youtube… and even on my site, I need to compromise between quality and bandwidth issues. The dvd thing was just something I threw together to play at a party. Vids will be all online via embed players. No plans for broadcast, or physical distribution as they’re all available for free.

[Chris Wright]

4:4:4 is chroma sampling
https://en.wikipedia.org/wiki/Chroma_subsampling

32bpc is levels of brightness
https://forums.creativecow.net/thread/2/917115

“So… 4:4:4 is synonymous with 32 bit color?”
No, you can have 4:4:4 8 bit or 4:4:4 16 bit.
GH2 output is 8bit. if you have a high end field recorder, there’s no point in trying to
capture 16 bpc. it’s like sending a value of 5 from the camera but you’re recording 5.000
but still try to match the 4:2:0 in the recorder. you don’t want to sub-sub sample your pixels!

“So, really, it sounds like I want to stick with 16 bit until final output?”
yes, just watch out for using 8bit effects, they will break the quality.

“Speaking of which, is there a way to look at an effect and know right off the bat if it’s processing in 8 or 16 bits?” Yes, in the effects panel, each effect has a small digit on the left, 8,16,32bpc

“C4d has an 8 bit dither checkbox for output. This should be off if rendering out in 16 bit?”
“”8 Bit Dithering is a process that adds a random pattern to colors to prevent color banding. Although dithering enhances the image quality, it also increases file size. For web graphics in particular, you may want to disable dithering to reduce image file size.As this “noises-up” the colors, the noise will be more prominent in the moment you need some heavy color correction.” So, basically render out of C4D in 16bpc if possible so you can color correct in AE.

also, if you’re exporting to trillions of colors, you are actually outputing 16bpc. Float is 32bpc. millions is 8bit. not all image seqences or codecs support all three. note:AE’s dither is different than C4D.

“Starting to see now why my first dvd experiment looked like a cartoon gone horribly wrong on the burned disc.”

Yes, besides losing a lot of quality in the mpeg2 from 4:2:0, the color profile needs to be color managed as well because dvd’s use rec.709 color management called an icc profile

color manage your monitor
https://www.johnpaulcaponigro.com/blog/tag/icc-profiles/

color management in After Effects
https://www.adobe.com/devnet/aftereffects/articles/color_management_workflow.html

“How exactly? What am I looking for?”
you’re looking for crappy looking quality 🙂

for youtube, they have their own recommendations for the type of file you upload.
https://support.google.com/youtube/answer/1722171?hl=en

for bandwidth issues, I’d definately look into adaptive video streams that can change quality
depending on the user’s connection type. Consider .flv files. they are very small.
but that’s a whole nother can of worms that could fill several books.