Encoding Script Optimized For YouTube

[dellsam34]

Any suggestions of a ffmpeg script to encode a lossless HuffYUV UHD file into a codec suitable for uploading into YouTube, So basically I just want to reduce the file size to stay below the YT file size limit, I would like to keep the 4:2:2 chroma sub and audio as PCM if possible, I've always uploaded lossless HD to YT since bandwidth is not an issue for me but 4k/UHD exceeds YT file size limit for uploading. I know YT compresses everything and makes it much worse, this is the exact reason I keep the file as lossless as possible before uploading so it gets butchered only once. A sample file is attached.

[davexnet]

What is youtubes bitrate limit ? THe only thing I see is a maximum 256 GB per file

[dellsam34]

My problem is not the bitrate, the file size is the limit.

[jagabo]

Something like (in a drag/drop batch file):

ffmpeg -y -i %1 -pix_fmt yuv422p10le ^
    -c:v libx265 -preset slow -profile:v main422-10 -crf 15 -colorspace bt709 ^
    -c:a copy "%~n1.x265.main422-10.mkv"

That delivered a 5.5 MiB file (from your 245MiB source).

[davexnet]

My problem is not the bitrate, the file size is the limit.

Well it's related.
a 30 minute file @ 1000kbs and a 15 minute file at 2000kbps would be about the same size...
so then the run time AND bit rate become a factor.

Your file is marked interlaced but it's progressive.
Encoded it in vdub2 using ffv1 lossless compression @ 422 chroma, PCM audio and received a file at about 55MBs
I

[jagabo]

The AVI file isn't flagged interlaced. MediaInfo always assumes huffyuv in AVI is interlaced.

[davexnet]

The AVI file isn't flagged interlaced. MediaInfo always assumes huffyuv in AVI is interlaced.

Good point.

[dellsam34]

I will try both ffmpeg and vdub variants and see which one looks better after an upload.

[dellsam34]

Something like (in a drag/drop batch file):

ffmpeg -y -i %1 -pix_fmt yuv422p10le ^
    -c:v libx265 -preset slow -profile:v main422-10 -crf 15 -colorspace bt709 ^
    -c:a copy "%~n1.x265.main422-10.mkv"

That delivered a 5.5 MiB file (from your 245MiB source).

How would I use this for a single file and where to insert the file name?

[davexnet]

I think jagabo's approach is better. Will give you the option to upload much longer files and shorter upload times.
I don't think it's necessary to keep it lossless, although on my PC, a bit on the slow side, the lossless encoding is quicker

[jagabo]

Something like (in a drag/drop batch file):

ffmpeg -y -i %1 -pix_fmt yuv422p10le ^
    -c:v libx265 -preset slow -profile:v main422-10 -crf 15 -colorspace bt709 ^
    -c:a copy "%~n1.x265.main422-10.mkv"

That delivered a 5.5 MiB file (from your 245MiB source).

How would I use this for a single file and where to insert the file name?

It's a drag-and-drop batch file. Put the code in a bat file (a plain text file with the extension .bat rather than .txt), then drag/drop a video file onto the bat file. Or put the bat file in your Send To folder. Then you can right click on a video file and select Send To -> (name of bat file) from the context menu.

Of you can type:

ffmpeg -y -i "input.avi" -pix_fmt yuv422p10le -c:v libx265 -preset slow -profile:v main422-10 -crf 15 -colorspace bt709 -c:a copy "output.mkv"

[dellsam34]

For some reason that -y caracter at the beginning of the the second script did not process the file, so I removed it and it worked, Here is both lossless and encoded files uploaded to YouTube:

Lossless: https://youtu.be/mGse2ZN94GQ
Encoded: https://youtu.be/h9j1KNmxnzA

[pandy]

If 256GiB is your problem then i need to ask what kind of length and bitrate are expected to be dealt... Personally if i need to upload something on YT then i prefer encoding speed instead file size (so i prefer preset rather fast/faster than medium/slow) - also i would probably avoid B frames at all (i use only I and P type - my observation leading me that there is less quantizer "pumping" visible in final render).

[Selur]

Small side question: Does it make sense to upload 4:2:2 content to YouTube, afaik. YouTube always converts to 4:2:0?

[jagabo]

Small side question: Does it make sense to upload 4:2:2 content to YouTube, afaik. YouTube always converts to 4:2:0?

I was wondering the same thing. Especially since the OP appears to be working with upscaled VHS.

Also, there's a rec.601 vs rec.709 color difference between the two youtube videos in post #12. The command line I gave explicitly flagged the video as rec.709. I assumed that because it was upscaled to 4K, though it's not flagged as such in the earlier huffyuv AVI.

For some reason that -y caracter at the beginning of the the second script did not process the file

That's weird because -y tells ffmpeg not to warn you if its output is going to overwrite an existing file, it just overwrites it. If there is no existing file the switch is ignored.

[dellsam34]

If 256GiB is your problem then i need to ask what kind of length and bitrate are expected to be dealt... Personally if i need to upload something on YT then i prefer encoding speed instead file size (so i prefer preset rather fast/faster than medium/slow) - also i would probably avoid B frames at all (i use only I and P type - my observation leading me that there is less quantizer "pumping" visible in final render).

Anywhere from minutes to hours depends on the material, For me quality is the priority, I would prefer to upload large file sizes because that doesn't bother me at all, No cap in data, upload speed is very high and I don't have to sit next to it to watch it uploading, The only limit is YT file size, hence encoding. One tape I'm uploading that initially exceeds YT file size limit, I was able to cut the size down using the FFV1 in vdub2 but it is only 41min, longer tapes will eventually go over.

Small side question: Does it make sense to upload 4:2:2 content to YouTube, afaik. YouTube always converts to 4:2:0?

Unless we know how YT algorithm handles both 4:2:2 and 4:2:0 as inputs, Sure everything is going to be mushed down, that's the whole reason. to keep the losses at minimum. The same way for resolution, did you think I want to upscale SD from VHS to 4K?

Also, there's a rec.601 vs rec.709 color difference between the two youtube videos in post #12. The command line I gave explicitly flagged the video as rec.709. I assumed that because it was upscaled to 4K, though it's not flagged as such in the earlier huffyuv AVI.

I do convert that to rec.709 in vdub output, Haven't used your script yet for an actual long file, Will try it later.

[pandy]

If 256GiB is your problem then i need to ask what kind of length and bitrate are expected to be dealt... Personally if i need to upload something on YT then i prefer encoding speed instead file size (so i prefer preset rather fast/faster than medium/slow) - also i would probably avoid B frames at all (i use only I and P type - my observation leading me that there is less quantizer "pumping" visible in final render).

Anywhere from minutes to hours depends on the material, For me quality is the priority, I would prefer to upload large file sizes because that doesn't bother me at all, No cap in data, upload speed is very high and I don't have to sit next to it to watch it uploading, The only limit is YT file size, hence encoding. One tape I'm uploading that initially exceeds YT file size limit, I was able to cut the size down using the FFV1 in vdub2 but it is only 41min, longer tapes will eventually go over.

Then if possible i would go with fixed QP (something reasonable like 8..16) and only I and P frames thats all - as Google will recompress this anyway then i would not bother trying to squeeze bitrate within refined preset - especially if you claim your upload speed is high (IMHO it can be justified to reduce filesize as much as possible even at a cost of longer encoding if upload speed will be slow but if your upload is fast then why bother to perform computationally expensive encoding) - quality is outcome of available bitrate and this is deciding factor - more sophisticated encoding tools are focusing on bitrate reduction even at a quality cost.

[jagabo]

Yes, if you don't need a lot of compression you can use a faster preset, a lower quantizer, and no b-frames.

Something like:

ffmpeg -y -i %1 -pix_fmt yuv422p10le ^
    -c:v libx265 -preset veryfast -profile:v main422-10 -crf 8 -x265-params bframes=0 -colorspace bt709 ^
    -c:a copy "%~n1.x265.main422-10.mkv"

The result was about 17.7 MB, compared to 5.5 MB for my earlier settings.

[pandy]

So math is simple 256GB means there is 2048 Gb available for file, this mean that let say file is 3600 seconds (1 hour) long then for 1 sec we have over 568Mb available so if your source HUFFYUV encoded was around 1400Mbps then H.265 need to compress source approx 2,5 times more efficient than HUFFYUV - let say HUFFYUV can provide 3 times compression so in overall HEVC need to be able compress source with factor 8 - this should be quite easy and it should provide quasi lossless compression. Of course this is quite simplistic estimation and for material longer than 1 hour compression must be more efficient but it should be not so difficult as in real life we expects compression factors at least one order higher than your case - probably 3 hour+ long source should be able to fit within 256GB limit.

[dellsam34]

All makes sense, Thank you guys.

[dellsam34]

This is the 138GB lossless version upload:

https://youtu.be/YWZsX_EGOK4

[pandy]

So YT offers your video at best with 16.4Mbps - there is long way from 448Mbps in your source.