I am trying to compress some of my video collections with ffmpeg, and I wish to accelerate it with my RX 6900XT.
For libx265, I know that I can use the following command:
ffmpeg -i input.mp4 -v:c libx265 -crf 24 output.mp4
However, the -crf
flag is ignored by hevc_amf encoder.
I tried ffmpeg -h encoder=hevc_amf
, and it seems that there are flags like -rc cqp
to set a constant quantization parameter, CQP, which behaves similar to CRF (it seems that CRF is still a little bit more advanced allowing variable QP during the process), but I don't know how to specify a reasonable value for that option, and I wish to know if there are better approaches that work similarly to the CRF option for libx265?
-quality <value>
(from 0 to 10) option, in addition to-rc cqp
. Why not experimenting by yourself about the values ? There's apparently no CRF-like mode, the GPU encoders are usually much less flexible than x264/x265