Repolode
Submit
0xSero

0xSero/video-edit

An agent skill and script kit that turns a raw macOS screen recording into a tight, redacted, upload-ready video with a verified render.

View on GitHub
Stars
85
Forks
4
Open issues
0
Last commit
3 days ago

What is it?

video-edit is a set of Python and ffmpeg scripts plus an agent skill for turning a 45–90 minute macOS screen recording (OBS, screen with webcam) into a 20–30 minute video ready for upload. The workflow is driven by a transcript: an agent decides which ranges to keep, and the scripts build an edit decision list, render it, and check the result. Everything runs locally.

The kit targets common problems in raw recordings: long stretches of visually static screens, which are detected by frame-difference analysis rather than silence; pauses that are shortened only where the enhanced audio is quiet; and sensitive text such as hostnames or API keys, which is covered with blur windows and then checked by OCR on the rendered output. Apple Vision is used for hand-pose and face detection to flag candidate segments for review. Audio goes through DeepFilterNet3 noise reduction, EQ, de-essing, compression and two-pass loudnorm to −16 LUFS.

Renders are restartable by group, and a verification script re-derives the expected frame count, compares audio track lengths, and re-runs OCR on the finished file, exiting non-zero on any leak. A selftest script runs the full pipeline on a synthesized recording.

Key features

  • 01Transcript-driven keep ranges and privacy cuts defined in a single job.json file.
  • 02Static-screen detection by frame difference, separate from silence detection.
  • 03Whole-pane blur for sensitive text, followed by OCR of the rendered output to confirm no leaks.
  • 04Audio chain with DeepFilterNet3 noise reduction and two-pass loudnorm to -16 LUFS.
  • 05Restartable grouped h264 rendering on macOS with VideoToolbox, plus a verification gate.

Use cases

  • Cutting a long OBS screen recording of a coding session or demo into a shorter YouTube video.
  • Redacting hostnames, tokens or other on-screen secrets before publishing a recording.
  • Removing idle stretches and tightening pauses in talks recorded on a Mac.

Quick start

  1. 1

    Clone the repository into your agent skills directory

    git clone git@github.com:0xSero/video-edit.git ~/.agents/skills/video-edit
  2. 2

    Install Python dependencies

    python3 -m pip install mlx-whisper deepfilternet pillow numpy soundfile pyobjc-framework-Vision pyobjc-framework-Quartz
  3. 3

    Check the install by running the self-test on a synthesized recording

    bash scripts/selftest.sh /tmp/ved-selftest

Things to know

  • Requires macOS 14+; Linux is not supported as-is.
  • Built for a specific OBS-style macOS screen recording workflow.
  • Requires ffmpeg 6.0+ and Python 3.10+ installed separately.

In collections

Open-source Mac appsPrivacy & local-first toolsPDF, OCR & document toolsAI video & audio tools

Topics

#video-editing#screen-recording#ffmpeg#audio-processing#privacy#ocr#macos#automation

FAQ

What is video-edit?
An agent skill and script kit that turns a raw macOS screen recording into a tight, redacted, upload-ready video with a verified render. video-edit is a set of Python and ffmpeg scripts plus an agent skill for turning a 45–90 minute macOS screen recording (OBS, screen with webcam) into a 20–30 minute video ready for upload. The workflow is driven by a transcript: an agent decides which ranges to keep, and the scripts build an edit decision list, render it, and check the result. Everything runs locally.
Is video-edit free to use? What license does it have?
video-edit is open source under the MIT license. Check the license terms before using it in commercial products.
Is video-edit still maintained?
Actively maintained. The last commit to video-edit was on Nov 4, 2026.
What can I use video-edit for?
Cutting a long OBS screen recording of a coding session or demo into a shorter YouTube video. Redacting hostnames, tokens or other on-screen secrets before publishing a recording. Removing idle stretches and tightening pauses in talks recorded on a Mac.
How do I install video-edit?
Clone the repository into your agent skills directory: git clone git@github.com:0xSero/video-edit.git ~/.agents/skills/video-edit Install Python dependencies: python3 -m pip install mlx-whisper deepfilternet pillow numpy soundfile pyobjc-framework-Vision pyobjc-framework-Quartz Check the install by running the self-test on a synthesized recording: bash scripts/selftest.sh /tmp/ved-selftest
Where can I get video-edit?
The source code and installation instructions are on GitHub: https://github.com/0xSero/video-edit

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects