Andrew, I believe the simplest solution, which I should have tried in the very beginning, is just to add these two lines to the mxf_avcintra_100.mxf file:

keyint_min=25
x264-params keyint=25

At least that worked for me and so far only 48000 audio samplereate has worked for any of the mxf render formats that I tested. I will check a modified mxf_avcintra_100.mxf file into GIT after I do more testing. Sorry for the extra work I have imposed on you.

On Fri, Jun 26, 2026 at 4:50 AM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:

I think my patch

"Do not set our options for mxf muxer"

is not needed really, because it was different set of errors and it only "worked" because I failed to properly change profiles during my testing on Termux? I'll keep it around just in case.

I tried to add g=1 to our DNxHR/dv mxf profiles too just in case, accidently added two .seg profiles I was testing some time ago (for segmented muxing)

Changed openjpeg mxf profile to use internal jpeg2000 encoder, libopenjpeg based one is seems to be gone from ffmpeg (note decoder may still be there)

Added g=25 to our mxf_h264_annex_b profile, may be it will work after this change?

Not really tested yet, too many machines (real and virt) trying to do many things at once.