Found this code on github via some comment discussion
ah, such a mix of opencv/qtgui!
but may be it exposes enough to make it into pure c++ pass?
Still no idea how dynamic metadata work .. there are standarts but they a bit hard to read for me:
Tonemapping using those already supported by ffmpeg's libplacebo filter, but this one need Vulkan driver/device!
And how to *construct* those types of metadata is not clear for me .....