Essential Information: Bitrate x Time = File Size. That's all that matters.

• All else being equal, the Bitrate and File Size of a true 10 bit encode (from 10 bit source!) will be 1/3 larger than its 8 bit counterpart.
• If you encode the two files at the same bitrate, the 10 bit file will have lower temporal quality, usually visible in motion areas and fades.
• An 8 bit file encoded in a 10 bit wrapper will be virtually the same size, because all you will have done is add air. Actual tests below.
• 

Ok Thank you for your explanations. That's all I wanted to know. Regards

@Musicvid Can you explain one more wonder for me?
What will have better/best quality as a final rendered video:
a) when I would record raw video in 4:2:2 and render as 4:2:0

or

b) when I would record raw video in 4:2:0 and render as 4:2:0

c) is it a sense to record raw video in 4:2:2 and render the result video as 4:2:0?

Thank you for your answer
Marcin

Some people prefer to grade from 4:2:2 source and render 4:2:0 for delivery, claiming it gives them more control. I was unable to confirm or deny those claims in controlled tests since downconversion involves a simple exclusion of data, although colorful, unsubstantiated stories abound on the internet.

I would welcome your own quantified (meaning "not" subjective) testing on the subject, since you seem interested in these things. Always glad to be proven wrong on a level playing field.

@Musicvid Ok Thank you for your answer. I do really apprieciate your knowledge.

For green screen work and tracking you want the highest color resolution available, at a consumer level, it's often 422 10bit. , A TV studio camera used for chroma key as a comparison may output 444 12bit .The bit depth is more important for most people's uses

While it may be subjective, I find that when shooting 4:2:2, I have far more control over my reds, they don't wash out as easy, even in the final 4:2:0 output. I find this especially noticeable when shooting a stage with a band on it that is washed in harsh red light. Red seems to clip easier if it's 4:2:0 native... though these results are also complicated by the fact that the 4:2:0 video i'm comparing against is 8 bit... but still, that shouldn't have an effect on one color appearing muddy/clipped before the others do.

What will have better/best quality as a final rendered video:
a) when I would record raw video in 4:2:2 and render as 4:2:0

or

b) when I would record raw video in 4:2:0 and render as 4:2:0

c) is it a sense to record raw video in 4:2:2 and render the result video as 4:2:0?

 

Thank you for your answer
Marcin

Are these all 10-bit or higher? Perhaps you can share the exact settings your camera is capable of. Are you talking about raw video (not compressed)?

While it may be subjective, I find that when shooting 4:2:2, I have far more control over my reds, they don't wash out as easy, even in the final 4:2:0 output. I find this especially noticeable when shooting a stage with a band on it that is washed in harsh red light. Red seems to clip easier if it's 4:2:0 native... though these results are also complicated by the fact that the 4:2:0 video i'm comparing against is 8 bit... but still, that shouldn't have an effect on one color appearing muddy/clipped before the others do.

That's an interesting observation. Sometime can you shoot a comparison for us to look at?

@Former user Do you mean that motion tracking better results (with less failures) we can achieve in 4:2:2 10bits color depth than in 4:2:0 8 bits color depth?

Are these all 10-bit or higher? Perhaps you can share the exact settings your camera is capable of. Are you talking about raw video (not compressed)?

I haven`t meant raw uncompressed video (ProRes RAW). I have meant i.e. XAVC or hevc codec recorded in 10 bits.

I think 10 bit is a far larger difference than 4:2:2 vs 4:2:0. If shooting log I'd shoot 10-bit, color correct, and then output 8-bit 4:2:0.

@RogerS Why did you write that you suggest me to output video in 8bit while having 10bit as a raw video? I thought that when I would record a video in 10 bit than I should output video in 10bits either. I agree that output video should be better 4:2:0, because I also noticed that my tv box doesn't playback smoothly 4:2:2 but my tv box playback smoothly 4:2:0.

Outputing video in 8bits while having 10 bits in raw video would be, I think, out of sense or maybe I don`t understand your goal. Please explain it to me. Thank in advance for your explaination.

I think you're pretty confused. With few exceptions video is delivered in 8-bit 4:2:0.

However if you shoot 8-bit 4:2:0 and attempt to make large color corrections, it doesn't have the necessary precision and you'll start to notice artifacts like banding and posterization. This is especially critical for log gamma files (such as Sony S-log 3 and S-log 2) where a huge dynamic range is stored. 8-bit just doesn't have the gradations of levels needed to avoid artifacts in gradients like skin and sky. Shoot S-log 3 of a blue sky in 8-bit mode and you'll see this quite clearly with dancing lines where there should be smooth blue tones. Note, you also need to use 32-bit full or video in Vegas to take advantage of high bit files.

Too complicated? Shoot Sony Cine2 or similar gammas and you can stick to 8-bit capture and delivery. I don't know what your camera is but if you've got an a7SIII/FX3/FX6, etc. it's capable of more than 8-bit capture and the S-log files are extremely flexible in post.

@Former user Do you mean that motion tracking better results (with less failures) we can achieve in 4:2:2 10bits color depth than in 4:2:0 8 bits color depth?

Motion tracking relies on points of high contrast, and it's high contrast where 444 (no color subsampling) is needed the most, or less sub sampling, so 422 instead of 420. 420 can cause blurred edges in this situation due to the need to interpolate color information from adjoining pixels. When tracking your object, if lighting or background colors change this can change and move the tracking edges as perceived by the tracker

Normally you would film in 422:10bit , edit in 422 10bit color grading/keying/tracking, and output to 420 8bit or 10bit depending on requirements. That would be the ideal. Make changes where required

@David-Rogers thank you for explanation, but I worry of the following thing. Will rendering video files recorded in 4:2:2 10 bit and output them in 4:2:0 8 bits, decrease the quality of the output footage?

"Will rendering video files recorded in 4:2:2 10 bit and output them in 4:2:0 8 bits, decrease the quality of the output footage?"

Yes. No one else is ever going to see what you captured for many reasons. Possibly, the primary reason is that mortals don't have 10 bit TVs.

thank you for explanation, but I worry of the following thing. Will rendering video files recorded in 4:2:2 10 bit and output them in 4:2:0 8 bits, decrease the quality of the output footage?

In practice, no. It increases the quality of the output footage compared to shooting 8 bit 4:2:0. The extra bit depth is useful at the editing stage, not the viewing stage, as John says there isn't a 10-bit viewing pathway anyway for standard video (this will change going forward with HDR video but doubt you are doing that).

If you've ever done photo editing it's standard to shoot raw (14 or 16-bit depth), do extensive processing, and then output 8 bit jpegs for viewing. Even for printing you'd normally output 8-bit tiffs. The point of high bit depth is flexibility to use that data to create the best possible 8-bit final image.

What will have better/best quality as a final rendered video:
a) when I would record raw video in 4:2:2 and render as 4:2:0

or

b) when I would record raw video in 4:2:0 and render as 4:2:0

c) is it a sense to record raw video in 4:2:2 and render the result video as 4:2:0?

@marcinzm Doesn't make any sense to me. But my 4k cameras only have single physical 4:2:0 sensors in them. Your probably does too. They have options to upscale in-camera to 4:2:2 or 4:4:4 but I don't see the point if I can record smaller and truer clips and upscale later in the computer or monitor. I do own a couple older triple-ccd hd cameras which actually capture 4:4:4 but downscale it to 4:2:2 in-camera. But they only record 8-bit which, although very good thanks to great lenses, their hd 4:2:2 avc is clearly inferior to 4k 4:2:0 10-bit hevc footage.

@RogerS, but let's say that I have 10bit TV set and my clients also have 10bit TV set. Let's stay with it that my clients doesn't have video knowledge and doesn't give me any video "must have" video preferences for the result video file they will get from me. I would like to give them the best quality video I am able to give them. Should I give them 10bits or maybe 8bits?

Downsampling 10 bit 422 to 8 bit 420 requires some form of dithering during encoding to prevent clumping and splotches.

Dithering means adding noise in some form to smooth and feather the chroma edges.

It is an arguable point that dithering noise more than negates any theoretical net advantage of grading in a higher bit domain. It is also provable that the inevitable frequency domain noise added to shadows as a natural consequence of downsampling is interpreted (incorrectly) by some observers to also mean a reduction in low-frequency banding.

My own tests are now completely invalid, as I believe that Vegas no longer uses pattern dithering during downsampling. I think it would be really nice if some of our younger hotshots would design there own empirical testing models, and enlighten us all, again.

Caution: testing uses more energy than talking.

but let's say that I have 10bit TV set and my clients also have 10bit TV set. Let's stay with it that my clients doesn't have video knowledge and doesn't give me any video "must have" video preferences for the result video file they will get from me. I would like to give them the best quality video I am able to give them. Should I give them 10bits or maybe 8bits?

I don't believe this theoretical set exists today. For standard dynamic range video, 8-bit 4:2:0 is the format to target. I don't think an increase to 10-bit or greater would even be visible for competently mastered video.

If your client asked for an ungraded video for their in-house colorist to finish, follow their specifications and it would likely mean delivering a 10-bit format with light compression like a type of ProRes.