Techniciens audio et vidéo Add accurate subtitles to videos from ASR or supplied subtitle files, producing a standalone SRT and matching transcript by default, with an optional burned-in video only on explicit request or post-delivery confirmation. Use when the user asks to 给视频加字幕, 生成字幕版视频, 重做字幕同步, add subtitles/captions, burn subtitles into a video, create fine-grained subtitles, or optionally add a horizontal dynamic module progress bar (which first delivers a YouTube/Bilibili chapters file). Requires fine-grained/word-level ASR rather than splitting coarse subtitles, semantic cue segmentation at phrase boundaries (not at raw audio-gap timing), slide-text (OCR) grounding for proper nouns and on-screen numbers, conservative proofreading, hallucination filtering, Simplified/Traditional Chinese script normalization, filler-word cleanup when requested, proper-noun/terminology review before final SRT delivery, and final SRT gate validation (plus ffprobe/frame validation when rendering).
2026-06-08