ffmpeg-video-audio-sync

First compare stream durations with ffprobe, then explicitly trim the longer stream using -t before combining rather than relying on -shortest. Also normalize audio to AAC 44100Hz stereo 192kbps to prevent sample-rate drift over time.

When all MP4 segments share start_time of 0.000000, using -f concat with -c copy causes PTS collisions that silently drop segments without errors. Convert each MP4 to TS format, binary concatenate with cat, then re-encode back to MP4.

Convert each MP4 to MPEG-TS with -c copy -f mpegts, concatenate the TS files with binary cat, then re-encode the combined TS to MP4. Direct cat on MP4 files fails due to the moov box structure.

Yes. Scenes encoded with -preset ultrafast or fast can lose roughly 25 percent of data during TS conversion with -c copy. Always encode source segments with -preset medium or higher before container conversion.

Use -stream_loop -1 on the video input combined with -shortest so the video repeats until the audio ends. This avoids truncating the narration when the video is shorter than the audio track.

Run ffprobe with -select_streams v:0 and a:0 to read each stream duration separately. A difference under 0.2 seconds is acceptable; also confirm total nb_frames equals the sum of all segment frame counts.