[FFmpeg-devel] Pixel format 12bit grayscale

[Reimar Döffinger]

On Mon, Apr 13, 2015 at 09:59:57AM +0000, Marco Porsch wrote:
> Previously we had regular 16bit RAW that worked like a charm as input to FFmpeg. The naïve approach of upscaling the 12bit to 16bit destroys the compression efficiency of course...

I think there's your problem, that is not a "of course at all".
Upscaling 12 to 16 bit should take a at best negligible amount of extra
encoding space with any sane compression method.
(besides the existing possibility as mentioned later of telling the
encoder how many of those 16 bits actually are relevant).