encoding for youtube

[marcorocchini]

Hi cats,

please consider my source is like this:

General
Complete name : S:\comedia\esp\test.avi
Format : AVI
Format/Info : Audio Video Interleave
Format profile : OpenDML
File size : 3.18 GiB
Duration : 32s 880ms
Overall bit rate : 830 Mbps

Video
ID : 0
Format : YUV
Codec ID : YUY2
Codec ID/Info : YUV 4:2:2 as for UYVY but with different component ordering within the u_int32 macropixel
Duration : 32s 880ms
Bit rate : 830 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate : 25.000 fps
Color space : YUV
Chroma subsampling : 4:2:2
Compression mode : Lossless
Bits/(Pixel*Frame) : 16.004
Stream size : 3.18 GiB (100%)

simply it a fullHD interlaced 4:2:2 source 25fps

Now I would like encode it for youtube: if I don't mistake youtube recode all, and recode in 4:2:0 colorspace.

Using ffmpeg I try to use a commandline like this:

ffmpeg.exe -y -i s:\comedia\esp\test.avi -i s:\comedia\esp\test.wav -pix_fmt yuv420p -vf yadif -c:v libx264 -profile:v main -level:v 4.1 -g 33 -bf 1 -acodec pcm_s16le -aspect 16:9 g:\testcommedia.mkv

I wonder if this commandline is suitable or not for youtube, or is necessary any other parameters. Thanks

[sneaker]

You probably want "yadif=1" to enable double frame-rate deinterlacing. Profile main, -g 33, -bf 1, level 4.1 are not necessary but you ought to increase quality/bitrate using e.g. -crf 14 unless your uploading speed is extremely low.

[marcorocchini]

consider also that I have to scale from my interlaced source 4:2:2 into 4:2:0 progressive so ffmpeg have to do a colorspace conversion, is there necessary a line something like

-vf scale=interl=0 (or 1?) ?

[SameSelf]

ffmpeg is one way to go about this. However, if you are:

(Note: Avisynth command equivalent added after the -->)
1. Dithering down to 8 bit-->e.g. LWLibavVideoSource(source="myclip.avi", format="YUV422P8")
2. Converting the colorspace-->ConvertToYV12()
3. De-interlacing--> YADIF() or highly recommended QTGMC()
4. Doubling frame rate-->ChangeFPS(), or if using in conjunction with a de-interlacer like QTGMC, nothing is needed
5. Rescaling-->Spline36Resize(), LanczosResize(), etc.
6. And so on...

That is a lot to pack into a ffmpeg command line and difficult to test. I recommend looking into Avisynth along with VirtualDub to do all that which allows you to test various settings/filters first as well as avail yourself the superior deinterlacing and scaling filters in Avisynth. Then you can frameserve the Avisynth script to ffmpeg or x264. The only thing left to worry about in your command line is the final encode parameters (e.g. bitrate, profile, etc.).

[_Al_]

His early scripts used Avisynth, he abandoned them because indexing supposedly takes too much time. If you ask me not worthy, Avisynth is highly versatile, could be replaced with Vapoursynth perhaps in the future for 10bit etc....

[pandy]

consider also that I have to scale from my interlaced source 4:2:2 into 4:2:0 progressive so ffmpeg have to do a colorspace conversion, is there necessary a line something like

-vf scale=interl=0 (or 1?) ?

Up to you - deinterlace first - assumption is that rest of processing is progressive.
Consider to upscale video to UHD if you need to preserve very fine details.
Don't waste time for fancy encoder presets - YT will re-compress anyway everything.
WEBM on YT offering higher quality so encoding to VP9 sound fair as alternative to H.264.

[SameSelf]

His early scripts used Avisynth, he abandoned them because indexing supposedly takes too much time. If you ask me not worthy, Avisynth is highly versatile, could be replaced with Vapoursynth perhaps in the future for 10bit etc....

I see. That is too bad. While yadif is ok, I always choose QTGMC if, for no other reason, it's multi-threaded, a must for any de-interlacing filter IMO. My understanding is that the UltraFast preset is identical to yadif. Thus for someone who is de-interlacing, I would think that would trump any slowness caused by indexing, especially for FHD content. I can push SD content through QTGMC MT at about 70 fps on my PC while HD content is around 20 fps IIRC. I would be curious to know what his fps encoding speed is and how much that would speed up or slow down if using QTGMC in MT mode on the UltraFast preset since speed seems to be a concern.

EDIT: I wonder if QTGMC in UltraFast runs in a MT mode since yadif is not multi-threaded? So maybe there is nothing to be gained?

[_Al_]

It is for youtube so yadif is perhaps more than enough.

Rest I don't know, I just did something for others ( I do not have interlace HD) deinterlacing and downresizing HD interlace, producing DVD or SD broadcast and one can get away with yadif doing that,especially with video with bad quality in the first place. I did not troubleshoot for thread use, MT etc.,

Is it really visible at the end watching yadif 1 or QTGMC on big screen, HD content?

For SD original content , deinterlacing SD using QTGMC is a must though.