kajisho5/ffmpeg-skill v3.6.5

on GitHub

Automated release: version and notes generated from pull requests merged since 3.6.4.

Fixed

• Caller text and JSON files are read one way, and odd input is read as ffmpeg reads it or
  refused as kind: input (#457, #465: A262, A269, A73, A19, A268, A16, A15, A270, A66, A12,
  A20, A14, A13, A18, A17). Text files go through one reader that drops a UTF-8 BOM and ends every
  line at \n (\r\r\n is one line end): metadata.py --chapters, silence.py --filler-words and --words and join.py --list no longer keep the BOM in their first entry
  (join's --list is also refused as input, not a traceback, when it is not UTF-8). SRT cues are
  grouped by timing line: an SRT with no blank lines between cues, \r\r\n line ends, position
  fields (X1:.. Y1:..) or a second --> reads as ffmpeg's SRT demuxer reads it, and a cue line of
  only spaces no longer splits the SRT write_srt() writes. caption.py --text splits a line at
  \n only (not U+2028 or U+0085), and reads 3 to 5 apples as text: with to both sides must be
  clock times (0:03 to 0:05); 0-1 Hello still reads bare seconds. JSON files a caller names
  (render projects and plans, batch recipes and their project, brand files, --words, cut.py --snap-source) nested too
  deep, or of the wrong type where the run reads them (clips[].speed, audio.stems,
  export.preset; a plan's inputs, unbound_inputs, verify, commands, cwd; a recipe's glob, project,
  output_dir, suffix, ext, clip_key, an absolute glob or a step that is not a list; a brand
  that is not an object or whose caption/styles.caption is not one), are refused as
  kind: input; a plan gets the new error.reason plan_field_invalid. In a whisper word-timing
  JSON, a segment of the wrong shape (a string or a list) contributes no word timings, as the
  reader's docstring says, while valid words elsewhere in the file are still read;
  silence.py --words refuses a file only when no valid word is left. caption.py --ass
  with --offset refuses a file that is not UTF-8 on the real run as the dry run does.
• fix: one reader for caller text and JSON files; SRT cues keyed on timing lines (#483)
• test(join): cut, crop and join of three H.264 + AAC parts keeps every frame (#482)
• test(join): short H.264 + AAC MP4 clips keep their frames (#481)

What's Changed

• test(join): short H.264 + AAC MP4 clips keep their frames (#475) by @kajisho5 in #481
• test(join): cut, crop and join of three H.264 + AAC parts keeps every frame (#479) by @kajisho5 in #482
• fix: one reader for caller text and JSON files; SRT cues keyed on timing lines (#457, #465) by @kajisho5 in #483

Full Changelog: v3.6.4...v3.6.5