kajisho5/ffmpeg-skill v3.6.6

on GitHub

Automated release: version and notes generated from pull requests merged since 3.6.5.

Fixed

• A tool writes the picture and the sound it measured, on a file with two of either (A110,
  A118, A128, A258). Where a command had no -map, FFmpeg's default selection chose the largest
  picture and the track with the default disposition or the most channels, while every
  measurement and check was taken from what probe() describes: the first picture that is not a
  cover (V:0) and the first audio track (a:0). loudness.py measured track 1 and applied that
  gain to track 0 (a -16 LUFS target came out more than 10 LU short); export.py exported a different track
  depending on whether the source had a subtitle stream; cut.py and crop.py wrote the larger
  picture they never probed (crop.py's rectangle was checked against the other one);
  stabilize.py analysed and wrote it too. audio.py --mono read the channel count of track 0
  for the --audio-stream track, so a stereo track came out stereo and a mono one lost 6 dB.
  _common.probe.probed_maps() now maps those streams, and the shared measurements
  (decode_pcm_mono, decode_gray_frames, measured_level_dbfs, detect_silences,
  detect_scenes, analyze_levels) read them, as do check.py's loudness row, cropdetect.py,
  look.py, scenes.py --sheet, silence.py's cut levels and sync.py's output. A file with
  one picture and one sound builds the same cut.py stream copy as before; with several, the copy
  maps the first subtitle stream where the container holds it as it is and reports it dropped
  (dropped_non_av_streams) where it does not.
• fix: map the picture and sound probe() describes on every tool (#484)

What's Changed

• fix: map the picture and sound probe() describes on every tool (A110, A118, A128, A258) by @kajisho5 in #484

Full Changelog: v3.6.5...v3.6.6