video-transcriptlisted

# Video Transcript Skill Use this skill when the user asks for a video transcript, video-to-text, YouTube transcript, or a Markdown/text稿 from a video URL or local media file. ## Core Policy 1. **Do not download full video just to get text.** Try caption/subtitle retrieval first. 2. **Use captions before ASR.** Captions are faster and usually cost nothing. 3. **Only download audio for fallback transcription.** If captions are missing/unusable, download the best audio stream and run local Whisper. 4. **Preserve provenance.** Record input URL/file, selected method, language/caption source when known, and output path. 5. **For Telegram or chat delivery, prefer a `.md` attachment for long transcripts.** Avoid dumping very long transcripts into one message. ## One-command workflow From this skill directory after `yk install video-transcript`: ```sh python3 scripts/video_transcript.py "https://www.youtube.com/watch?v=..." \ --format markdown \ --output /absolute/path/transcript.md ``` For a local VTT/SRT caption file: ```sh python3 scripts/video_transcript.py /absolute/path/captions.vtt --format markdown ``` For a local audio/video file that needs ASR: ```sh python3 scripts/video_transcript.py /absolute/path/talk.mp4 \ --format markdown \ --output /absolute/path/talk-transcript.md ``` Useful options: - `--langs "zh-Hans,zh-CN,en,ja,ko"` — caption language priority for `yt-dlp`. - `--no-asr` — captions only; fail instead of downloading audio/transcribing. - `--br