So, Should I edit (color grading etc.) and work my project in 8 bits, and then, just before make my final render, change it to 32-bit (video levels)? Right?

I'm creating my projects in 32-bits pixel format (video levels) to prevent levels clipping between filters in chain (not to prevent banding).

If one of your filter moves colour outside of 0-255 level, it might be possible to return that colour back to allowed range using another filter in chain.

You can follow these steps to see it in action.

Some caveats:

- Working with projects in 32-bits pixel format is noticeable slower

- If you already applied come levels/colour corrections you would better not to change project's pixel format as you can get wrong results. Check all your corrections if you switched project's pixel format for some reason.

- Some filters does not work with 32-bits format

- Some filters (Color Curves for example) by its nature produce the same effect as clipping and this cannot be reversed because they are NOT moving colour outside of valid range

- It was a bug in VP15 when 32-bits pixel format projects were rendered incorrectly using SonyAVC or XDCAM-EX encoders (fixed in build 361). Magix AVC you mentioned in other thread is fine, however it has another issues and not acceptable for me :)

Grading in 32 bit has no effect at all on banding in 8 bit source, where the damage has already been done.

I did a round of tests and Vegas' pattern dither negates the already negligible effects of grading at higher bit depth. IOW, I was unable to determine any difference to delivering 8 bit door to door.

but Vegas currently lacks the ability to properly color grade 10 bit or higher video, so what you see while editing won't be what you get when you render)

Which is precisely the reason not to load 8 bit source in a float project. It's always been that way, frosty, at least since 2.0. And it can't be fixed with intervention before rendering, not completely.

I understand it is on the "fix" list, maybe already been implemented?

keep in mind the human eye can't see 16.8 million colors so when placed side by side, an 8-bit version and a 16-bit version of an identical image will look = identical to us. Also if the original is 8 bit you can't pump out a 16 bit render and get better than the 8 bit. so where is the benefit then? Flexibility or Elasticity in the tonal values of your edit so since you have 8 bit bicycle adding another wheel will mean you have to carry it while trying to steer. just go with the 8 bit and enjoy the ride.

This is all not about seeing all these colours.

This is mostly about preventing unnecessary degradation during video processing.

... that can't be seen in tests.

its connected = linked like a marriage if you only understand or focus on the one side you fail to see other side then you fail overall. the post is about banding caused by an 8-bit image pushed out of its range. an 8-bit image, which gives us 16.8 million possible colours . so do you need more?  Working in 32 bit with an 8 bit image won’t give you more of what you don’t need.  AND CAN’T GET. is the point

https://www.photoshopessentials.com/essentials/16-bit/

So for example here is a 8 bit clip normal from my camera then I pushed the levels in 8 bit SO if I work in 32 bit video levels and output = same.   

... that can't be seen in tests.

Really? Cannot you see that one colour (from 3) is disappeared in demo I suggested?

the post is about banding caused by an 8-bit image pushed out of its range.

Not only. The post about reasons to process 8-bits footage in 32-bits projects. Not only banding happens here.

https://blogs.adobe.com/VideoRoad/2010/06/understanding_color_processing.html

Anyway, it is up to you to decide what options would you like to use. I do not use 8-bits project as it might cause quality lose, I do not use GPU encoders ad they produce worse results with the same bitrate etc. Trade of is speed. I'm aware of this.

@fifonik

Yes, your "demo" works as expected. And I do applaud your effort in providing a better example than the typical "1+1=2, therefore do as I say" kind of logic we usually get around here.

However, I don't see the REAL WORLD application of this demo. Why would you need to blow your highlights into outer space and then bring them back?

So again I ask: PLEASE provide an image sample from real 8 bit camera footage that shows obvious image improvement due to 8 to 32 back to 8 processing, versus 8 to 8 to 8 processing. I have an open mind, but cannot find a situation where this applies.

I might have add Color Correction filter to track and increase saturation (1.2). In this case some bright colours will be "moved out of valid 8-bit range". Mathematically it is something like c2 = c1 * 1.2. c can only hold 1 byte (0-255). So if c1 == 220 it would become 264 and truncated to 255.

Then I might add Levels filter to fragment so everything are in 16-235. However, with 8-bits format I will not be able to do this. Everything that was more that 212 already truncated to 255 so 212, 213, ... 255 would look the same after that levels adjustments. With 32-bts project I can.

So again I ask: PLEASE provide an image sample from real 8 bit camera footage that shows obvious image improvement due to 8 to 32 back to 8 processing, versus 8 to 8 to 8 processing. I have an open mind, but cannot find a situation where this applies

This is not possible. You cannot improve your 8-bit original image. It is only possible to make the image processing less destructive.

... or more destructive, neither of which you seem able to quantify. Good story, though.

Well, generally I like to think I am improving my clips by colour grading / correction, so "less destructive" is not the goal. But I get your point - 32 bit is a "safety net" of sorts.

By the way, are you sure the track fx would be applied before the event fx? (I think you mean event when you refer to "fragment"). This does raise the point that order of operations is important and not all of these mistakes are corrected by using 32 bit. For example, applying sharpening fx to film grain fx.

> Well, generally I like to think I am improving my clips by colour grading / correction

Sure! This is why people doing this! It definitely looks better after correction.

I meant that 32-bits processing cannot magically convert 8-bit footage to something better.

> By the way, are you sure the track fx would be applied before the event fx?

Sure track fx applied after event fx. Sorry.

Sure, effects must not be applied in random order and 32 bits is not a silver bullet. Main idea was -- 32 bits never worse than 8-bits and preventing truncating/clipping while data transferred from one filter in chain to another.

> But I get your point - 32 bit is a "safety net" of sorts

Yup. That's it.

Main idea was -- 32 bits never worse than 8-bits

Never? Sometimes? Always?

In the presence of identifiable gamma mismatch, subsequent downsampling degradation, and complete lack of visual evidence? Don't be ridiculous, you are tossing confetti.

"The use of absolutes never engenders credibility in the absence of supporting, verifiable data." By who? Beginning to catch on now?

But not to worry. I catch myself doing it too.

Difference between 8bit and 32bit-floating point (full Range)

First is 8bit (no plugin)

Two is 32bit (full range) (no plugin)

- Compositing Gama: 1.0

- ACES version: 1.0

- ACES color space: Default

- View transform: sRGB (ACES)

If the video is dark I use it Vegas Level plugin.

The side by side images need to have same levels for comparison. Changing to full range obviously changes the levels. Please try 32 bit video levels.