What problem does it solve? Raw ASR transcripts contain wrong proper nouns, homophone errors, and awkward line breaks that make spoken-video subtitles hard to read. This Skill turns a video or audio file into a polished, proofread subtitle file aligned with the picture. ## Core Features & Use Cases - Speech Transcription: Extracts 16k mono audio and runs ASR (SenseVoice, Whisper, or Volcengine) to produce timestamped results. - Spoken-Style Segmentation: Breaks lines by breath and semantics rather than written punctuation, keeping each screen readable (roughly 15 Chinese characters per line). - AI Proofreading: Corrects proper nouns via a terminology glossary, fixes homophones, and normalizes numbers and punctuation. - Output & Burn-in: Generates SRT or styled ASS files and optionally burns subtitles into the video with ffmpeg (soft or hard subtitles). - Use Case: After editing a talking-head video, feed the final cut in and receive a proofread subtitles.srt plus a hard-subtitled export ready for publishing. ## Quick Start Transcribe this video, proofread the subtitles against my glossary, and generate an SRT file plus a hard-subtitled version.
