kajisho5/ffmpeg-skill v2.20.0 on GitHub

Automated release: version and notes generated from pull requests merged since 2.19.0.

join.py reports the clips' own total beside the joined length. The expected length a join is checked against comes from the transition the join applied, so a fade that took time out of the material passed unseen. 15 s of clips joined with the default fade play as 14 s. inputs_duration and transition_shortening now sit next to expected_duration, the duration step carries inputs_total, and notes says when the transitions shortened the join.
silence.py says which removed spans were digital silence. Quiet sound under --threshold (room tone, a breath, a whisper) and digital silence were removed alike and both reported as silences. They still are, but silence_kinds labels each span digital or below_threshold, and silence_breakdown gives the count and seconds of each.

loudness.py no longer lengthens the sound. Through the AAC encoder its output ran longer than the input: 0.015 s on a 6 s clip, 0.1 s on a 90 s one. The sound is trimmed to the input's length, and a duration step measures it.
export.py measures the length its preset promises. It trimmed to the platform maximum but never re-read what it wrote. With --normalize, a Reels export left at 90.10 s by the loudness pass came back verified: true, and check.py failed the same file. A duration step now holds the written file to the same limit check.py uses.
Lengths and silences are checked against a yardstick the tool did not make (#372)

What's Changed

Lengths and silences are checked against a yardstick the tool did not make by @kajisho5 in #372

Full Changelog: v2.19.0...v2.20.0