kajisho5/ffmpeg-skill v2.19.0 on GitHub

Automated release: version and notes generated from pull requests merged since 2.18.0.

Traditional Chinese captions get a Traditional face. Han-only text always resolved a Simplified Chinese font, so a Taiwanese or Hong Kong caption was drawn in the mainland glyph forms (骨, 直 and 角 differ). --lang zh-TW, zh-HK, zh-Hant or yue on caption.py and graphics.py (or a brand's lang) now resolves a Traditional face: Noto Sans CJK TC, or Microsoft JhengHei on Windows. A machine without one draws with the Simplified face and records font_variant_missing.
Filler words in Korean, Chinese, Russian, Dutch, Polish and Turkish. silence.py --filler has built-in lists for ko, zh, ru, nl, pl and tr, so a transcript in one of them is no longer refused. The lists hold hesitation sounds only (음/어, 嗯/呃, э/эм, ehm, yyy, ıı). Words that are as often content stay out and are named as discourse markers for --filler-extra: 저 is also "I", 那个 is also "that". ISO 639-2 codes (kor, zho) and language names (chinese, spanish) in a --words document now name the right list; spanish used to be cut to sp.

Japanese fillers split across whisper tokens are cut. Whisper splits Japanese and Chinese into tokens, not words: えーと arrives as えー + と. Only a single token could match, so --filler never cut えーと or えっと and cut えー with と left behind. A run of up to four tokens that joins to a listed word now matches as that word. In Chinese and Japanese a match also needs punctuation or a 0.1 s pause on each side, so the あの of あの人 ("that person") is kept.
--filler-extra "you know" matches. A two-word marker could never match one engine word, though --help offered it.
whisper.cpp transcripts in Japanese and Chinese have per-token timings. whisper.cpp marks a word start with a space, so in an unspaced language a whole segment was glued into one "word" that no filler could match. Tokens between two whole characters now stay apart.
Filler words in Korean, Chinese, Russian, Dutch, Polish and Turkish; Japanese fillers split across whisper tokens; Traditional Chinese fonts (#371)

What's Changed

Filler words in Korean, Chinese, Russian, Dutch, Polish and Turkish; Japanese fillers split across whisper tokens; Traditional Chinese fonts by @kajisho5 in #371

Full Changelog: v2.18.0...v2.19.0