first or all: great project. I don't know what I would do without it... It puts the default encoding capabilities in various video editing software to shame...
As to my issue:
I have an Nvidia 1080Ti which in higher than 8 bit color depth not only supports YUV 4:2:0 hardware encoding but also 4:4:4.
ffmpeg.exe -h encoder=hevc_nvenc
>Supported pixel formats: yuv420p nv12 p010le yuv444p p016le yuv444p16le bgr0 rgb0 cuda d3d11
When using ffmpeg from command line I get an output that has 4:4:4.
That means that even if I have 4:2:2 12bit footage like yuv444p12le nvenc will auto-switch to the higher pixel-format without quality loss:
ffmpeg.exe -i in.mov -c:v hevc_nvenc -preset p7 -tune hq -rc vbr -multipass:v fullres -qmin:v 15 -qmax:v 20 -pix_fmt yuv444p12le -an "out.mkv"
>Incompatible pixel format 'yuv444p12le' for codec 'hevc_nvenc', auto-selecting format 'yuv444p16le'
I realize that it will cost more disk space in such an Incompatible pixel format event, but at least I do not loose any quality.
But when I select HEVC (NVIDIA NVENC) In Voukoder the only thing I can select is 8bit or 10bit. Not the chroma subsampling. Not 12bit.
Not the pixel format itself. Then I end up with a 4:2:0 encoded 10bit video, loosing both quality in color depth as well as chroma subsampling.
When selecting CPU encoding there is a detailed dropown for the pixel format (4:2:0 10bit, 4:2:2 10bit, 4:4:4 10bit and so on) - but encoding on the CPU takes ages......
Would it be possible to add support for that?
If not a specific chroma subsampling option (as there might be lots of combinations depending on the hardware)
Maybe at least an option to specify pix_fm manually as filter parameter?
Thanks a lot in advance!