SubMiner/changes/japanese-subtitle-generation-docs.md

type: docs area: subtitles

• Explain how dialogue generation retains uncertain audible sections, why songs may also be transcribed, how detected speech starts guide long-passage cuts to reduce early subtitles, and why each passage uses a fresh Whisper process.
• Document automatic loaded subtitle references, English dialogue preference, signs/songs filtering, fallback behavior, and the limits of reference-guided timing.