More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    clip video

    1

    Convert long-form videos into captioned vertical shorts with face-tracking and social metadata, running 100% locally.

    $14.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    clip video

    clip video

    Example session with this skill installed

    I need 5 clips from this podcast https://www.youtube.com/watch?v=dQw4w9WgXcQ for a TikTok and Shorts campaign. Use the channel name for the output folder.

    • Read your context and instructions
    • Compiled the clip video

    I have processed the video and generated 5 high-potential clips. The pipeline ran locally on your Mac, utilizing mlx-whisper for transcription and OpenCV for face-tracking.

    Clip Selection & Metadata

    IDDurationHook TitleMoment Summary
    0142sThe Secret to ConsistencyHigh-energy opening discussing daily routines.
    0235sWhy Most Startups FailCritical insight into market timing and team structure.
    0358sMy $1M MistakeEmotional reveal about a failed product launch.
    0428sThe 2-Hour RuleActionable productivity hack for deep work.
    0545sWhere Tech is HeadingForward-looking prediction on local AI models.

    All clips have been rendered to ./clips/podcast-channel-shorts/ at 1080x1920 resolution with word-by-word captions.

    Sample Metadata (Clip 03)

    TikTok Caption:
    The moment I realized I lost $1,000,000. 📉 This was the hardest lesson of my career but the most important. If you're building a product right now, listen to the 45-second mark.

    Hashtags: #entrepreneur #startup #businesslessons #founder #failureislearning

    Next steps

    1. Open CLIPS.md in the output folder to copy the formatted descriptions for each platform.
    2. Review the contact sheets in work/review/ to ensure the face-tracking crop didn't cut off any speakers.
    3. Listen to the audio transitions in Clip 03, as it was merged from two segments.

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

    About this skill

    Repurposing long-form video into high-engagement shorts usually requires expensive SaaS subscriptions and cloud processing. This skill provides a complete, 100% local pipeline for macOS to transform YouTube links or local files into vertical clips without API costs or data leaving your machine.

    What it does

    • Automated clipping uses local transcription to identify viral moments, hooks, and payoffs within long videos.
    • Face-tracking crop applies a smart 9:16 vertical crop that follows speakers, automatically switching to "fit" layouts for wide scenes.
    • On-screen captions burns word-by-word animated subtitles using Poppins ExtraBold and hooks using Anton typography.
    • Social metadata generation writes platform-specific titles, descriptions, and hashtags for TikTok, Reels, and YouTube Shorts.
    • Multi-part editing allows cutting dead air and joining non-contiguous segments into a single cohesive clip.

    How it works

    1. Fetch and process via yt-dlp and ffmpeg to extract high-quality video and 16kHz audio.
    2. Local transcription using mlx-whisper on Apple Silicon to generate word-level timestamps.
    3. LLM selection where you review the transcript to pick moments based on standing alone and strong hooks.
    4. Render and verify using the provided Python scripts to track faces, burn captions, and generate quality-control contact sheets.

    Frameworks & tools

    Built for macOS on Apple Silicon. Utilizes yt-dlp, ffmpeg, uv, and a Python environment running mlx-whisper, OpenCV, and numpy.

    Why this beats prompting it yourself

    Manual prompting cannot handle the heavy lifting of face-tracking, video encoding, or frame-accurate caption syncing. This skill bridges the gap between LLM reasoning and professional video editing by providing the specific rendering engine and local ML models needed to produce finished files.

    Use cases

    • Podcast repurposing to turn hour-long interviews into 60-second highlight reels.
    • Creator campaigns for participating in official clipping programs with specific rule enforcement.
    • Stream highlights for extracting the best reactions and stunts from gaming or live sessions.
    • Local-first production for creators who want to keep original footage and transcripts off third-party servers.

    Known limitations

    Requires Apple Silicon (M-series) for MLX acceleration. Transcription and rendering are resource-intensive. User must manually verify audio joins and crop accuracy via contact sheets.

    Requires macOS on Apple Silicon. First run downloads ~1.6 GB model. Cannot hear audio; user must verify joins and pronunciation manually.

    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 26 days ago

    Requires

    Python
    macOS

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables

    File Scopes

    ~/.cache/clip-video/**
    ./clips/**
    /tmp/clip-video/**
    Listed26 days ago

    What's inside

    Frequently Asked Questions

    This skill runs 100% locally on your machine, eliminating monthly SaaS fees and ensuring your video content and transcripts never leave your hardware.

    The skill is optimized for macOS with Apple Silicon (M-series chips) to leverage MLX acceleration for high-speed transcription and rendering.

    Yes, it uses a smart face-tracking algorithm to center speakers and automatically switches to 'fit' layouts for wide scenes or non-human subjects.

    You will receive a complete local pipeline including the Python rendering engine, pre-configured environment scripts via 'uv', and the integration logic for MLX-Whisper and FFmpeg.

    The skill generates platform-optimized titles, descriptions, and hashtags for TikTok, Instagram Reels, and YouTube Shorts alongside the finished video file.

    While the transcription and moment selection are highly automated, users should review the generated contact sheets to verify crop accuracy and audio transitions before final export.

    New in Image & Video