kajisho5/ffmpeg-skill v3.8.1

on GitHub

Automated release: version and notes generated from pull requests merged since 3.8.0.

Fixed

• An input that states no length is refused, not read as 0 s long (#491). A Matroska written
  to a pipe has no duration in its container or its streams (probe.py reports duration: null).
  reverse.py, grid.py and broll.py ended in a traceback with no JSON document; fit.py
  refused a "speed factor 0.00x", freeze.py and speedramp.py a clip of 0..0.000,
  silence.py "nothing would be kept"; join.py joined the clip as 0.04 s and fit.py --method trim, scenes.py and metadata.py --auto-chapters completed on a length of 0. The tools that
  compute with the length (fit --duration, freeze, speedramp, reverse, loop, silence,
  scenes, look's sheet, graphics without --end, metadata chapters, export --max-mb,
  join, grid, broll, multicam's reference, --fix-drift) read it through one helper,
  input_length(), which refuses as kind: input ("the length of FILE could not be read") with
  the hint to remux (export.py INPUT --preset copy). A tool that does not need the length runs
  on the file as before.
• grid.py, broll.py and multicam.py end their sound to the sample (#493). They trimmed
  it with atrim=duration={:.3f} / atrim=0:{:.3f} and the output's -t {:.3f}, to the
  millisecond, so a length of 59244 samples (48 kHz) was written as 59232. They share
  audio.py's conform now (conform_to_length() in _common: apad=whole_len=N, atrim=end_sample=N at a pinned rate, then a trim on the timestamp so sound that starts late
  still ends with the picture); audio.py's own filter strings are unchanged. The sound is
  padded with silence when it ends before the picture, and grid.py conforms every
  --audio-from track, not only a padded one. A dry run's pending input keeps the old filter.
  An MP4's stated stream duration is still a whole millisecond (its edit list is in 1/1000 s);
  the track holds the samples.
• fix: an input that states no length is refused, not read as 0 s; grid, broll and multicam end their sound to the sample (#497)

What's Changed

• fix: an input that states no length is refused, not read as 0 s; grid, broll and multicam end their sound to the sample by @kajisho5 in #497

Full Changelog: v3.8.0...v3.8.1