YouTube Transcribe

Transcribe YouTube videos with smart fallback: extracts captions first (fast, free), falls back to local Whisper transcription when no captions available. Au...

Security Scan

The skill does what it says: it extracts captions via yt-dlp and falls back to local Whisper transcription; its requirements and instructions align with that purpose and it doesn't request unrelated secrets or suspicious installs.

Purpose & Capability

Name/description match the implementation: the code uses yt-dlp to extract captions and to download audio, then attempts local Whisper backends (mlx, faster-whisper, openai-whisper). Declared required binaries (python3, yt-dlp) and optional ffmpeg are appropriate and sufficient for the stated functionality.

Instruction Scope

SKILL.md instructs running the included Python script with YouTube URLs and documents env var overrides. The script only reads system memory info (sysctl or /proc/meminfo), uses temporary directories, invokes yt-dlp/ffmpeg subprocesses, parses VTT files, and imports local Whisper backends. It does not reference unrelated files, external endpoints beyond YouTube/yt-dlp, or request arbitrary environment secrets.

Install Mechanism

Install specs are brew formulas for yt-dlp and ffmpeg (standard package sources). The skill does not auto-download code from an arbitrary URL or include high-risk install steps; Whisper backends are mentioned as pip installs but not automatically fetched by the skill itself.

Credentials

No credentials or sensitive environment variables are required. Two optional env vars (YT_WHISPER_BACKEND, YT_WHISPER_MODEL) are reasonable and documented. The script reads only common system info (RAM) to pick model sizes.

Persistence & Privilege

The skill is not always-enabled, does not modify other skills or global agent settings, and does not request persistent elevated privileges. It runs as a user-level script and uses temporary files.

Assessment

This skill appears internally consistent and implements the advertised behavior. Before installing, consider: 1) it will download YouTube metadata and audio to a temporary directory (yt-dlp/ffmpeg are used), so only run it with links you trust and where downloading is permitted; 2) Whisper backends and models (if used) can be large and may need GPU/CPU resources—you must pip-install the backend you want and may incur large model downloads; 3) the script runs yt-dlp subprocesses (no shell injection vulnerability visible because arguments are passed as lists), but you should still review the included script if you have a high security posture; 4) there are no requested secrets or outbound endpoints beyond standard YouTube/yt-dlp usage. If any of the above is unacceptable (downloading media, installing models, or running subprocesses), do not install the skill.

Smart YouTube video transcription with automatic fallback:

1. Captions first — extracts existing subtitles (manual or auto-generated) via yt-dlp. Fast, free, no compute.
2. Whisper fallback — when no captions exist, downloads audio and transcribes locally with the best available Whisper backend.

When to Use

Use this skill when the user wants to:

• Get a transcript or text version of a YouTube video
• Understand what a YouTube video says without watching it
• Summarize, analyze, or take notes from a YouTube video
• Extract subtitles or captions from a video

Triggers

• "transcribe this YouTube video"
• "what does this video say"
• "get the transcript of [YouTube URL]"
• "summarize this YouTube video" (transcribe first, then process)
• Any YouTube URL shared with a request to understand its content

Requirements

Required:

• yt-dlp — for caption extraction and audio download
• python3

For Whisper fallback (when no captions available):

• ffmpeg — for audio processing
• One of these Whisper backends (auto-detected in priority order):
  1. mlx-whisper — Apple Silicon native, fastest on Mac (pip install mlx-whisper)
  2. faster-whisper — CTranslate2 backend, fast on CUDA/CPU (pip install faster-whisper)
  3. openai-whisper — Original Whisper, universal fallback (pip install openai-whisper)

Usage

Basic — transcribe a video

python3 {baseDir}/scripts/transcribe.py "https://www.youtube.com/watch?v=VIDEO_ID"

Specify language for captions

python3 {baseDir}/scripts/transcribe.py "URL" --language zh

Force Whisper (skip caption check)

python3 {baseDir}/scripts/transcribe.py "URL" --force-whisper

JSON output

python3 {baseDir}/scripts/transcribe.py "URL" --format json

Save to file

python3 {baseDir}/scripts/transcribe.py "URL" --output transcript.txt

Options

Flag	Default	Description
--language	auto	Preferred subtitle/transcription language (e.g. zh, en, ja)
--format	text	Output format: text, json, srt, vtt
--output	stdout	Save transcript to file
--force-whisper	false	Skip caption extraction, go straight to Whisper
--backend	auto	Whisper backend: auto, mlx, faster-whisper, whisper
--model	auto	Whisper model size: auto, large-v3, medium, small, base, tiny

Environment Variables

Variable	Description
YT_WHISPER_BACKEND	Override Whisper backend selection
YT_WHISPER_MODEL	Override Whisper model size

Auto-Detection

Whisper Backend (priority order)

1. MLX Whisper — detected via import mlx_whisper. Best for Apple Silicon.
2. faster-whisper — detected via import faster_whisper. Best for CUDA GPU, good on CPU.
3. OpenAI Whisper — detected via import whisper. Universal fallback.

Model Size (based on available RAM)

RAM	Model	VRAM/RAM Usage
≥16GB	large-v3	~6-10GB
≥8GB	medium	~5GB
≥4GB	small	~2.5GB
<4GB	base	~1.5GB

Caption Language Priority

When --language is not specified, captions are searched in this order:

1. Video's original language
2. Chinese variants: zh-Hant, zh-Hans, zh-TW, zh-CN, zh
3. English: en
4. Any available language

Output Formats

text (default)

Plain text transcript, one continuous block.

json

{
  "video_id": "ZSnYlbIYpjs",
  "title": "Video Title",
  "channel": "Channel Name",
  "duration": 708,
  "language": "zh",
  "method": "captions",
  "transcript": [
    {"start": 0.0, "end": 5.2, "text": "..."},
    ...
  ],
  "full_text": "Complete transcript as single string"
}

srt / vtt

Standard subtitle formats with timestamps.