Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    yt2script

    1

    Turns a YouTube link into an English transcript, entirely on your machine. 將yt影片連結轉成繁中、簡中、英文逐字稿,完全地端運行,簡單好用。/yt2script + link

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    yt2script

    yt2script

    Example session with this skill installed

    Transcribe this talk for me: https://www.youtube.com/watch?v=dQw4w9WgXcQ

    • Read your context and instructions
    • Compiled the yt2script

    I have processed the video and generated two transcript files. The extraction used human-authored English subtitles (Step 1), ensuring 100% cue fidelity with no machine translation.

    [影片] Never Gonna Give You Up | 03:32 | Rick Astley (字幕 client=ios, 媒體 client=web)
    [探路] 人工字幕 1 軌 | 自動字幕 0 軌 | 影片語言 en
    [取得] 第 1 步 人工字幕 en
    [驗證] 54/54 cue 無損 ✓
    [完成] ./Never_Gonna_Give_You_Up.md

    Transcript Preview

    TimestampSpeaker/Content
    00:00:00[Music]
    00:00:18We're no strangers to love
    00:00:22You know the rules and so do I
    00:00:27A full commitment's what I'm thinking of
    00:00:31You wouldn't get this from any other guy
    00:00:35I just wanna tell you how I'm feeling
    00:00:39Gotta make you understand
    00:00:43Never gonna give you up, never gonna let you down

    Next steps

    1. View the full prose version in Never_Gonna_Give_You_Up.md.
    2. Use the timestamped version Never_Gonna_Give_You_Up(時間戳版).md for video editing or referencing.
    3. Ask me to summarize or translate this transcript now that the verbatim text is captured.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Transcribing YouTube videos often results in machine-translated junk, broken formatting, or hidden API costs. This skill provides a deterministic, local-first pipeline to extract verbatim transcripts without cloud fees or "hallucinated" summaries.

    What it does

    • Multi-source extraction prioritizes human-authored subtitles before falling back to native auto-captions or local Whisper inference.
    • Local Whisper processing uses large-v3 for high-fidelity speech-to-text on videos lacking any subtitle tracks.
    • Smart de-duplication removes scrolling caption artifacts and "roll-up" repetitions common in YouTube auto-generated text.
    • Native language detection ensures an English video yields an English transcript, preventing accidental machine translation into the user's locale.
    • Traditional Chinese optimization handles zh-TW conversion natively for Chinese audio sources.
    • Lossless verification checks every source cue against the final output to ensure zero data loss during processing.

    How it works

    1. Initialize environment by running the setup script to install yt-dlp, faster-whisper, and opencc in a localized venv.
    2. Fetch source data using fetch.py <URL> to probe for the best available text track or download audio for local inference.
    3. Build transcripts by running build.py on the temporary work directory to generate formatted Markdown files.
    4. Deliver artifacts including a clean prose version and a timestamped version, both saved to your specified directory.

    Frameworks & tools

    Built on yt-dlp for robust video probing, faster-whisper for local SOTA transcription, and OpenCC for Chinese script conversion. Runs via Python 3.11+.

    Why this beats prompting it yourself

    Standard LLM prompts often summarize or "clean up" speech, losing the raw fidelity of the source. This skill preserves every stutter, bracketed sound note, and pause, while handling technical hurdles like YouTube's "bot" detection and machine-translation flags that simple scrapers miss.

    Use cases

    • Convert long-form lectures or interviews into searchable Markdown archives.
    • Extract verbatim Traditional Chinese transcripts from Cantonese or Mandarin videos.
    • Recover text from videos with "burned-in" or missing captions using local GPU/CPU compute.
    • Generate high-quality datasets for RAG applications without incurring per-minute API costs.

    Known limitations

    Requires Python 3.11+. Initial Whisper model download is approximately 3GB. Local inference on CPU takes roughly 1.7x the video duration.

    Requires Python 3.11+. First-run Whisper model download is ~3GB. CPU inference takes ~1.7x video duration. Cannot bypass IP-based bot blocks via proxies.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 6 days ago

    Requires

    macOS
    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables
    Listed6 days ago

    What's inside

    Frequently Asked Questions

    Unlike basic scrapers, yt2script prioritizes human-authored captions and uses local Whisper large-v3 inference for videos without subtitles, ensuring you get verbatim text rather than machine-translated or summarized versions.

    This is a local-first skill that runs on your hardware using Python 3.11+; it does not require external API keys or subscriptions to services like OpenAI or Google Cloud.

    The skill includes a dedicated setup script that automatically builds a virtual environment and installs necessary dependencies like yt-dlp, faster-whisper, and OpenCC for you.

    Yes, it features a native Traditional Chinese optimization via OpenCC, specifically designed to handle zh-TW conversion for Cantonese or Mandarin audio sources.

    The skill is compatible with any AI agent framework capable of executing Python scripts, and it outputs clean Markdown files that are ready for RAG (Retrieval-Augmented Generation) applications.

    While it can run on a CPU (taking roughly 1.7x the video duration), the initial setup requires about 3GB of disk space to download the high-fidelity Whisper large-v3 model.

    New in Social Media