Proof point → When a tool earns its keep, I package it for the public: tests, CI, changelog, security policy, PyPI. That’s the difference between scripts and intellectual property.

🩻 Problem

A production line for music and content needs quality gates: releases that meet loudness standards, beat previews that can’t be stolen, shorts with burned subtitles, branded documents. Manual QC doesn’t scale; cloud tools leak unreleased work.

🔨 Solution

Six local-first tools, with the two 2026 flagships packaged to full open-source standard:

Architecture Overview

  1. rms-scan (audio_rms-detector) - wraps FFmpeg astats + EBU R128 to report RMS/LUFS and pass/fail against a configurable loudness window, with JSON output and exit codes designed for CI gating. Ships with tests, CI and release workflows, changelog, contributing/security/conduct policies - and a PyPI release.
  2. audio-watermarker - embeds spoken watermark tags at intervals with loudness-relative targeting based on local RMS (audible but not destructive), fades, panning, dry-run placement preview, plus BPM/key analysis and metadata/cover-art commands. Proper src-layout package with three test modules and dependency extras.
  3. video-clipping - a YouTube-shorts pipeline: Whisper and Vosk transcription, subtitle burning, clipping, and branding modules.
  4. Supporting utilities - wav→waveform rendering, a Pillow/PyMuPDF PDF stamper with positioning/opacity/page selection, and an honestly-attributed Three.js/GLSL audio-reactive visualizer adaptation.

📜 Philosophy

Local-first (unreleased music never leaves the machine), automation-friendly (JSON, exit codes, dry runs), and graduated: scripts stay scripts until they earn full packaging - then they get all of it.

🎓 Key learnings

  • Audio engineering in code: RMS vs LUFS, EBU R128, loudness-relative overlay mixing, BPM/key detection.
  • Python packaging done properly - pyproject, extras, console scripts, src layout, automated PyPI releases.
  • CLI design for pipelines, where the consumer is another program.

📈 Output & impact

  • A published PyPI package (rms-scan) with release automation - public, reusable IP.
  • A complete self-serve media-ops line supporting a working creative business.

🌍 Why this matters

Original Tools & IP. Turning private tooling into citable, distributable intellectual property - properly packaged, released, and licensed - is how builders and ecosystems compound their work. Not advocacy: practice, with the release workflows to show for it.

🚀 Hire me

Have tooling worth packaging into real open source? Let’s talk → · See also: RedditorJS · The thesis