What transcode setting should I set here?

--A wrote on 2/14/2019, 3:40 PM

Probably I need to transcode my files.
I need to know what setting give me the same result of quality of my clips.

Here are my clips specification:

Format : MPEG-4 at 524 Mb/s
Length : 3,08 GB for 50s 517 ms

Video #0 : JPEG at 522 Mb/s
Aspect : 4096 x 2160 (1.896) at 29.970 fps

Audio #0 : PCM at 1.536 kb/s
Infos : 2 kanały, 48,0 kHz

Can you tell me what setting should I set to got losless quality of transcoded clips?

Here is what can I choose...

Like you see i have not 522 Mb/s, i have only 140 Mb/s usable. IS NOT THIS TOO SMALL?

Comments

Marco. wrote on 2/14/2019, 5:10 PM

The 522 Mbit/s is based on an outdated uneffective I-frame-only-like CBR compression method. AVC is a modern, effective GOP and VBR based compression method. 140 Mbit/s is a lot for AVC encoding and should be superior over 522 Mbit/s M-JPEG.

But you might not be on the right way. If you really need mathematical lossless encoding you should not use AVC.
If you're fine with visually lossless encoding you might be fine with 140 Mbit/s AVC encoding (and remember your Quicktime M-JPEG video is lossy encoded, too).

--A wrote on 2/14/2019, 7:28 PM

Thanks you, Marco. That will help a lot.
But tell me...

If you really need mathematical lossless encoding you should not use AVC

What method FROM MY LIST above can I use instead?

Marco. wrote on 2/15/2019, 2:03 AM

If you are looking for a mathematically lossless transcoding you could try "AVI uncompressed" (if "Nieskompresonwany" means "uncompressed"). Test it with a single clip and see what the file size will be …

--A wrote on 2/15/2019, 9:43 AM

If you are looking for a mathematically lossless transcoding you could try "AVI uncompressed" (if "Nieskompresonwany" means "uncompressed"). Test it with a single clip and see what the file size will be …

Excatly. "Nieskompresowany" means "uncompressed"/ You are smart :)
But all we know that Uncompressed AVI there will be way toooooo large file.

Marco. I am looking for COMPRESSED kind of codec to choose but loseless :)

Marco. wrote on 2/15/2019, 10:43 AM

I can't say if there is a compressed but mathematically lossless encoding listed in that screenshot. The video codecs listed there doesn't offer it, in my opinion. And to the main containers listed there is no info about its codecs, e. g. there are several compressing lossless codecs usable inside the AVI container, but you probably would need to install these codecs first.

A good choice for a lossless compressed codec is MagicYUV RGB. You would use it with an AVI container and it was intensively tested and rated here. but it's not (or no more) free available ($ 9,00).

But why not just using a visually lossless codec for transcoding, like millions of video pros do? If there is no visible loss, why caring about it?

--A wrote on 2/16/2019, 3:06 PM

y not just using a visually lossless codec for transcoding, like millions of video pros do? If there is no visible loss, why caring about it?

I am scary about quality of my video. If I no longer can import files directly from Canon 5D Mark IV, I have to find another solutionm. One of my solution was to transcode to Mp4. But...
Subsampling was no longer that good.
In original Canon 5D Mark IV 4K files I had a 4:2:2, but after transcode to my new MP4 with bitrate 140 Mb/s I only got 4:2:0 subsampling. It is a terrible loss for for example color grading.
Marco, I have to know a new, the best soultion for my future work with these files. So I am asking what codec should I use & that codec can't make my source videos worse.

I also did a few test. Thanks for advice of MagicYUV codec and link.
What is interesint, I have got a 1820 MB/s from original clips 500 MB/s. Isn't this strange?

Also I wonder why after transcode to MagicYUV I can't check a subsumpling value anymore.
Now, after transcode my MediaInfo software doesn't show me if that clips are 4:2:2 or 4:2:0. Do you know why and how to understand this? I need to know that.

Marco. wrote on 2/16/2019, 5:02 PM

"In original Canon 5D Mark IV 4K files I had a 4:2:2, but after transcode to my new MP4 with bitrate 140 Mb/s I only got 4:2:0 subsampling. It is a terrible loss for for example color grading."

I don't know which transcoding tool you used. The FFmpeg batch file I provided in the other discussion transcodes these Canon clips to 4:2:2.

"What is interesint, I have got a 1820 MB/s from original clips 500 MB/s. Isn't this strange?"

This is exactly what I would have expected. Like I already mentioned several times, the codec your Canon camera uses for recording is (rather) lossy compressed. MagicYUV is mathematically lossless. Thus the file size difference.
It's the price to pay to have mathematically lossless against visually lossless. I'm not aware of any mathematically lossless encoder which outputs smaller file sizes than MagicYUV does (see the discussion I linked above).
If you want smaller file sizes you can't use mathematically lossless transcoding. If you want mathematically lossless transcoding you have to accept big file sizes.

"Now, after transcode my MediaInfo software doesn't show me if that clips are 4:2:2 or 4:2:0."

If you used the compression option "RGB" (which is recommended for a real lossless compression), the subsampling is neither 4:2:0 nor 4:2:2. It's 4:4:4 then. RGB is available as 4:4:4 or 4:4:4:4 (RGBA) only.

--A wrote on 2/18/2019, 1:40 PM

Marco, thanks for response.

It looks like I have two method of doing this right now.

1st, transcode via ffmpeg (code from you)
2nd, transcode via my transcoder with MagicYUV codec

Both of them will give me great compression results.

Both of them will not be less then 4:2:0 after transcode, right?

Marco. wrote on 2/18/2019, 1:55 PM

"Both of them will not be less then 4:2:0 after transcode, right?"

MagicYUV encoder used with RGB option is 4:4:4.

The FFmpeg script I published here maintains the color sampling of the source file, so if your source is 4:2:0, the rendered file will be 4:2:0. If your source file is 4:2:2, the rendered file will be 4:2:2. You could recheck the FFmpeg rendered file by a MediaInfo analyze.

Also note, in the FFmpeg script there is a CRF value ("-crf 20"). This CRF value determines the quality where lower values means better quality. Values in the range of 12 to 18 are considered to be visually lossless (and usually no one uses values below 12).
So if you decide to use that FFmpeg script but also want to tweak the resulting quality a bit more, try lower CRF values down to about 12 (though probably anything below 16 might be a waste of data rate).

Eagle Six wrote on 2/18/2019, 6:23 PM

@--A has a tool you can download to make comparisons between original and render versions, which provides a mathematical quality comparison. As an example I compared a DNxHD 4:2:2 10 bit source media to a rendered version using Magic YUV RGB. The results showed Magix YUV RBG to have zero quality loss. The source media had a bit rate of 440 MB/s while Magix YUV had 690. The source media sample file size was 628 MiB, while the Magic YUV file was 972. In comparison to uncompressed, the unpressed render file was also lossless, but the 2,983 MB/s and the file size turned out to be 4,090 MiB. As you can see there is a huge difference between uncompressed and Magix YUV RGB, yet Magix YUV RGB is also lossless and way more quality than the source from your Canon 5D. So you are not going to lose any quality with the Magic YUV RGB and the increase in file size will be minimal.