boring-tools

Use when you have a long take and want local, text-based cuts, captions, or vertical and landscape social or ad renders, with every decision stored in a readable file that can be rebuilt.

Provide the raw footage and plain-language direction. You do not have to prepare the cut list, timeline, or rebuild script in advance. Documented tools are ffmpeg, whisper, and python on macOS, with Homebrew packages, a pinned requirements file, and a separate model download. Claude Code is described as working in the terminal or desktop app, not the browser. Those tools and models were not installed or run here.

The qualified post is a same-timestamp reply that walks through a local install and the editing workflow; the parent teaser is context only and was not given a second card. The teaser video and the linked YouTube episode were not watched. After setup, place the footage on the machine, give direction in ordinary language, keep every cut and caption decision in readable files, rebuild from those files, and report the verify numbers. Do not fetch models during a video job. Documented commands and scripts were not executed here.

The documented host is macOS, with Apple Silicon used for the default hardware encoder; that was not tested here. Keep runtime offline, and do not execute a remote repository during a video job. The README claim that the tools do not upload footage, face, voice, or transcript, and the post wording that the video never gets uploaded, are source statements, not an independent test. Self-reported twelve-file results, the review-grid image, reporting-shape numbers, and unwatched media are not verified examples. On-screen product claims must be checkable. MIT requires notices and gives no warranty. No source skill, script, or media was executed or inspected.