On Wed, 8 Apr 2020 at 19:04, Andy Sheimo <[email protected]> wrote: > This still produces the same error; > > ffmpeg -threads 1 -loglevel verbose -hwaccel cuvid > -hwaccel_output_format cuda -c:v mpeg2_cuvid -extra_hw_frames 3 -i > "l:\welter_originals\gamera the brave\Latatude Zero-trailer.mkv" -c:a copy > -c:v h264_nvenc test.mkv > > But this works to complete the job; > > ffmpeg -threads 1 -loglevel verbose -hwaccel cuda > -hwaccel_output_format cuda -extra_hw_frames 3 -i > "l:\welter_originals\gamera the brave\Latatude Zero-trailer.mkv" -c:a copy > -c:v h264_nvenc test.mkv > > By not specifying -c:v mpeg2_cuvid do we know what decoder is being used? > Is the process still completely in hardware? Do you think this is a > limitation of the cuvid decoder? > > Thanks for your help. -Andy >
NVDEC (an alias to the CUDA hwaccel) handles GPU memory quite differently from CUVID, and when its' in use, the decoder is not specified. CUVID, however, was never a true hwaccel, but rather a "wrapper" that "brings up"/initializes a specific decoder by name passed to it via -c:v, in a similar fashion to the bring-up seen with QuickSync (-hwaccel qsv). I cannot speak much about why CUVID is failing on your end. I migrated all my workflows to NVDEC an year ago. Perhaps omitting the hwacccel_output_format with CUVID will help? Try this: ffmpeg -threads 1 -loglevel verbose -hwaccel cuvid -c:v mpeg2_cuvid -extra_hw_frames 3 -i "l:\welter_originals\gamera the brave\Latatude Zero-trailer.mkv" -c:a copy -c:v h264_nvenc test.mkv Then report back. _______________________________________________ ffmpeg-user mailing list [email protected] https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email [email protected] with subject "unsubscribe".
