I need to point to some accurate places (chapters) on a video created with ffmpeg, in various formats (flv and mp4/h264).
But I can only point to the closest keyframe.
This works just fine with cuepoints created with video post prod software, though cuepoints are just metadata. So I assume the software creates a keyframe at the exact same time as the cuepoint to get such accuracy.
Unfortunately, I didn't find out how to create cuepoints, or keyframes, with ffmpeg. I only could set global parameters for keyframes, as the frames gap between 2 keyframes. But I can't set 10 keyframes per second just to get the needed accuracy for 3 chapters on a 20 minutes video...
Does anyone know how I could do this, with ffmpeg or, even, with any linux command line tool ?