For now it reads, if my copy is correct: cat bin/ffmpeg/video/prores_4444xq.pro mov prores_ks # ProRes, encoder prores_ks. Container Apple QuickTime movie (.mov). #Support for alpha channel profile=5 quant_mat=auto cin_pix_fmt=yuv444p10le #also gbrp10le # The possible video profile values for the ProRes codec are: # 5 = 4444xq # 4 = 4444 # 3 = HQ (high quality) # 2 = Standard # 1 = LT (light) # 0 = Proxy ===== but shouldn't line cin_pix_fmt=yuv444p10le reads like in prores_4444.pro cat bin/ffmpeg/video/prores_4444.pro mov prores_ks # ProRes, encoder prores_ks. Container Apple QuickTime movie (.mov). # Support for aplpha channel profile=4 quant_mat=auto cin_pix_fmt=yuva444p10le #also gbrp10le # The possible video profile values for the ProRes codec are: # 5 = 4444xq # 4 = 4444 # 3 = HQ (high quality) # 2 = Standard # 1 = LT (light) # 0 = Proxy Note that it uses yuv*A*444p, not just yuv444p ! Oh, and aplpha word was underlined with red wavy line in gmail editor - typo detected! currently playing with ffmpeg.git and this merge request: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/22418#issuecomment-32848 it seems that new Vulkan encoder can do 12bit even with alpha. But for some reason it works inside cingg but not with ffmpeg standalone! Note that this merge request does not work for me fully (unlike vanilla ffmpeg.git at 9c63742425aa9c84148f218513f74dbab1e01926) even in cingg, or more accurately it makes encoded video with some block artefacts/colored lines. Not for full footage, but for some parts of it.