Keeping 4:2:2 all the way to final encode?

[latreche34]

Is it a good idea to keep 4:2:2 all the way to the final encoded file?

I capture lossless 4:2:2 and de-interlace using QTGMC then encode to H.264, Previously, I use to down sample to 4:2:0 in the encode stage but one of the members told me that I can keep the 4:2:2 in the final file using this commend:

ffmpeg -i Input.avi -vf "format=yuv422p,setsar=sar=8/9" -c:v libx264 -crf 10 -x264opts colorprim=smpte170m:transfer=smpte170m:colormatrix=smpte170m:force-cfr -c:a aac -b:a 192k Output.mp4

Here are some figures of file sizes:
A lossless sample is 290MB.
After being de-interlaced it yields a file of 1.16GB (This is just a temporary step)
Encoding to H.264/420 gives 65MB
Encoding to H.264/422 gives only 79.6MB, not a big difference.

Now I wonder if it's a good idea to set the audio to 16bit/48Khz from 24bit/48Khz rather than a lossy audio?

What do you guys think?

[lordsmurf]

For about 5 years now, I routinely
- capture lossless 4:2:2,
- often encode out a 4:2:2 15mbps+ MPEG-2 archive copy,
- and deinterlace a copy for CRF 15-20 H.264 to 4:2:2

As noted, the difference in file size is small.

Lossless audio really is not needed, just use a high bitrate to encode (160kbps+ AAC 48kHz stereo, 384kbps AC3/MP2 aduio for MPEG). You can encode audio as high as you want, but at some point it's just wasted bits and file bloat. Audio isn't video.

I only leave 4:2:2 for the disc/streaming specs (4:2:0).

Never 4:1:1, far too lossy. (I don't care what theory is, practical application is very definitely lossy, even to non-video laymen.)

[latreche34]

I figured the audio is not going to be a big issue, I'm actually encoding at 192Kbps, I just thought maybe for pre recorded video clips a PCM 16/48 might be a good choice but I was told for H.264 will be some format compatibility issues not file size issues. the SAR is actually 10/11 not 8/9, it was copy/paste mistake.

[lordsmurf]

H.264 doesn't pair with PCM in MP4/MKV wrappers.

I only have PCM on the lossless masters, for later restoration. By the time it sees H.264 (or often MPEG), it's already been restored (if needed). The main issue is having a high enough bitrate that doesn't alter ambient sounds (hiss, etc) to digital noise. Sometimes, like when hiss is present, I just go insane on the audio bitrate. 448kbps, probably more than double needed, just in case.

But 4:2:2 from start to end? Yep, very sound.

After prodding from sanlyn and some others at VH, I've even started to deinterlace out to 59.94/50fps. Again, it's not a huge file size change. Some years ago, there were reasons to keep 29.97 as deinterlaced 29.97 -- and noting that 29.97 still doesn't really "throw away" much/any data (as the frames are all rebuilt). But with bandwidths and HDDs as they are now, 59.94/50 is maximum SD long with Full D1 720x480 (which is also technically overkill for VHS). It may give better motion at times, maybe, so why not?

You go even more overkill, doing 10-bit SDI processing on a signal that is 6-bit dithered at best.

Crank it to 11.

No reason not to, when not constrained by specs.

[latreche34]

Most after capturing apps are not 10bit compatible, I'm forced to capture in 8bit. Don't know why though?