youtube-transcribelisted

# YouTube / Audio Transcription Turn a YouTube video — or any local audio/video file — into a transcript without relying on YouTube-provided captions. Many videos (especially re-uploads and some Chinese-language uploads) have **no** subtitles and **no** auto-captions, so downloading captions with `yt-dlp --write-subs` returns nothing. This skill sidesteps that by downloading the audio and transcribing it locally with a whisper engine chosen by platform. ## When to reach for this - The user gives a YouTube URL and wants the 字幕 / 逐字稿 / transcript / captions. - `yt-dlp --list-subs` reports `has no subtitles` / `has no automatic captions`. - The user has a local `.m4a`/`.mp3`/`.mp4`/etc. file to transcribe. ## Prerequisites - `uv` installed (used to run `yt-dlp` and the transcription script without polluting the global env). The script auto-selects the whisper engine and `uv` installs only the one your platform needs: - **Apple Silicon Mac** → `mlx-whisper` (MLX/Metal, GPU-accelerated). - **Intel Mac / Linux / Windows** → `faster-whisper` (CTranslate2, CPU or NVIDIA CUDA). - A JavaScript runtime — `node`, `bun`, or `deno` — on PATH. Recent yt-dlp needs one to extract YouTube; the script auto-detects whichever is present. - ffmpeg recommended but **not** required: if absent, the script falls back to macOS's built-in `afconvert` (so on non-macOS, install ffmpeg). ## Usage The whole pipeline is one self-contained script. Run it with `uv run` so dependencies are provisione