Youtube Whisper

Download YouTube videos and transcribe audio using local Whisper. Use when you need to extract text from YouTube videos that don't have subtitles, or when yo...

Security Scan

The skill does what it says (download a YouTube video and transcribe locally with Whisper); the included shell script and SKILL.md are coherent with that purpose, though the registry metadata omitted declaring required binaries.

Purpose & Capability

Name/description match the actual behavior: the script downloads audio (yt-dlp), optionally extracts subtitles, and runs a local Whisper CLI to transcribe. However, the registry metadata lists no required binaries or environment variables while the SKILL.md and script clearly require yt-dlp, ffmpeg and a 'whisper' CLI (openai-whisper). This mismatch is a packaging/manifest inconsistency that should be corrected.

Instruction Scope

Runtime instructions and the included script stay within the stated purpose: checking for subtitles, downloading audio to /tmp, running Whisper to transcribe, and writing transcripts. The script inspects system state (memory, CPU) to decide whether to proceed and prints the video title. One operational caveat: the script contains an interactive prompt when available RAM is low (read -p), which can block non-interactive/autonomous agent runs.

Install Mechanism

No install spec (instruction-only) — lowest install risk. The skill includes a shell script file but does not download or execute remote installers itself. Dependencies are expected to be installed on the host (yt-dlp, ffmpeg, openai-whisper).

Credentials

The skill does not request any environment variables, credentials, or config paths and the script does not access secrets; it only runs local system commands and network calls to YouTube via yt-dlp, which is proportionate to the stated functionality.

Persistence & Privilege

The skill is not forced-always, does not modify other skills or system configuration, and only writes temporary files to /tmp and the chosen transcript output path. No elevated or persistent privileges are requested.

Assessment

This skill appears to do what it claims, but check a few things before installing/using it: (1) The manifest omitted declaring required binaries — you must install yt-dlp, ffmpeg, and a Whisper CLI (e.g., pip install openai-whisper which provides the 'whisper' command) before running. (2) The script is interactive when memory is low (prompts with read -p) and can block non-interactive/autonomous runs — run manually or modify the script for non-interactive use. (3) The script downloads content to /tmp and calls yt-dlp (network access to YouTube); ensure you are comfortable with downloading the target videos and that doing so complies with copyright/YouTube terms. (4) As with any third-party shell script, inspect the file and run it in a controlled environment or sandbox if you have any doubt. (5) If you plan to publish or automate this skill, ask the author to update registry metadata to declare required binaries and clarify the non-interactive behavior.

YouTube Whisper / YouTube Whisper 語音轉文字

下載 YouTube 影片並使用本地 Whisper 進行語音轉文字。

硬體需求 / Hardware Requirements

項目	最低需求	建議需求
記憶體	4 GB	8 GB+
儲存空間	1 GB	5 GB+
CPU	任意	Apple Silicon M 系列

測試環境 / Test Environment

項目	規格
主機	Mac mini M4
記憶體	16 GB
模型	Whisper small
影片時長	5:55
轉錄時間	約 3-5 分鐘

When to Use / 使用時機

• YouTube 影片沒有字幕 / YouTube video has no subtitles
• youtube-watcher 因為沒有字幕而失敗 / youtube-watcher fails due to missing captions
• 需要離線轉錄 / Need offline transcription
• 高品質本地轉錄 (比 YouTube 自動字幕更好) / High-quality local transcription

Usage / 使用方式

# 基本轉錄 / Basic transcription
youtube-whisper.sh "https://www.youtube.com/watch?v=VIDEO_ID"

# 指定輸出檔案 / Specify output filename
youtube-whisper.sh "URL" "output.txt"

# 指定模型 / Use specific Whisper model (tiny, base, small, medium, large)
# 預設語言: 繁體中文 / Default language: Traditional Chinese (zh)
youtube-whisper.sh "URL" "output.txt" "small"

Requirements / 需求

• yt-dlp: brew install yt-dlp
• Whisper: pip3 install openai-whisper or use openai-whisper skill
• ffmpeg: brew install ffmpeg

Output / 輸出

轉錄文字檔案 (Transcript text file).

Notes / 備註

• 影片會下載到 /tmp 並在轉錄後刪除 / Videos are downloaded to /tmp and cleaned up after transcription
• 預設模型: small (速度與準確度的最佳平衡) / Default model: small
• 較大的模型 (medium, large) 較慢但更準確 / Larger models are slower but more accurate

Author

Kuanlin