github kajisho5/ffmpeg-skill v4.3.0

one hour ago

Automated release: version and notes generated from pull requests merged since 4.2.2.

  • A wrong call is answered with the corrected command. Where every mistake on the line has one answer, the usage error ends with try: and the whole command, ready to copy: denoise.py --strength strong gets --strength high, silence.py --threshold -30dB --min-silence 0.8s gets --threshold=-30 --min-silence 0.8, cut.py --strat 1 gets --start 1. Nothing is suggested when a mistake has no single answer.
  • Three things most models asked for. proxy.py --height 360 (the width follows), redact.py --rect 100:40:180:60 (the rectangle in one value) and metadata.py --chapter 0:00=Intro --chapter 1:30=Main (chapters without a file). Each came from most of nine models given the same tasks blind.
  • tests/first_try.py gives a blind call a two-minute clip and real files for the names it invents (b.mp4, music.mp3, a.srt, logo.png), so a refusal is about the call. The strongest model's calls that run on the first try read 48 of 126 with it (39 before).
  • feat: the corrected command after a wrong call, and three options most models asked for (#656)

What's Changed

  • feat: the corrected command after a wrong call, and three options most models asked for by @kajisho5 in #656

Full Changelog: v4.2.2...v4.3.0

Don't miss a new ffmpeg-skill release

NewReleases is sending notifications on new releases.