В сообщении от Monday 09 November 2020 11:23:37 Andrea paz via Cin написал(а):
Yes, you are right, ffmpeg does not handle 12-bit DNxHR. Thanks. The options it offers (yuv422p; yuv422p10le; yuv444p10le and gbrp10le) are the only ones possible and therefore they are the right ones. I enclose the changes. The only problem is in using DNxHR_444: if you set the pixel format yuv444p10le the resulting file is RGB; if you set gbrp10le the resulting file is YUV. So the opposite happens! If you can verify this, please do me a favor.
Strange, it works fine for me - but I clicked on 'video options / view' and checked that right profile was selected. (so, 'profile ...' line jumped up one line in its window.) In another Cin in asset info I see: 1 video stream vid0 (0), id 0x000063: video1 dnxhd 1920x1080 30.00 pix gbrp10le for DNxHR-gbr10.mov and 1 video stream vid0 (0), id 0x000063: video1 dnxhd 1920x1080 30.00 pix yuv444p10le for /dev/shm/DNxHR-yuv10.mov so, for me there is no inversion, but may be because I clicked on options and not run it out of the box. No, it works even if I don't visit those settings ... BUT yuv422p10le and yuv422p doesn't work for me... in DNxHR_444.mov profile even if they are selectable .... [dnxhd @ 0xd0690b00] pixel format is incompatible with DNxHD profile [dnxhd @ 0xd0690700] ff_frame_thread_encoder_init failed FFMPEG::open_encoder err: Operation not permitted int FFMPEG::open_encoder(const char *, const char *): open failed dnxhd:/dev/shm/DNxHR-yuv10.mov Render::render_single: Session finished. [dnxhd @ 0xd348fc00] pixel format is incompatible with DNxHD profile [dnxhd @ 0xd348f840] ff_frame_thread_encoder_init failed FFMPEG::open_encoder err: Operation not permitted int FFMPEG::open_encoder(const char *, const char *): open failed dnxhd:/dev/shm/DNxHR-yuv10.mov Render::render_single: Session finished. this is with ffmpeg git, not 4.3 (while I can't see much changes in this codec between those?)
I renamed the audio preset DNxHR_pcm_s24.mov to DNxHR_pcm_s32.mov because in reality the only option it offers is 32-bit sampling and not 24-bit sampling.