This Skill Edited This Video. It Cost Less Than a Dollar.

A human editor charges about $40 per video. This skill runs for cents in API calls and does not need you in the room.

It takes the raw recording and the script, times every word to the audio, decides every cut, builds the caption track, picks sound effects and a music bed, and renders a finished 1080x1920 file. The finished MP4 drops into your Drive folder when it is done.

Two inputs in. One finished video out.

The Problem

Paying an editor per video is not the most expensive part of video production. The expensive part is being the bottleneck.

If you post three times a week, you have a dependency on another person's schedule for every piece of content you want to publish. The machine does not have an availability problem. It does not need style notes after the first run. And it costs less per video than a cup of coffee.

Beyond content: the same pattern applies to any business output that follows a repeatable template. Weekly reports. Client summary emails. Onboarding sequences. You are paying a human to execute the same steps in the same order every time. That is exactly what a skill is for.

What the Skill Does

Here is the sequence when you run it.

Step one: word alignment. The skill reads the audio against the script using speech recognition. Every word in the script gets a timestamp. This map drives every cut decision that follows.

Step two: visual placement. You provide a list of links: screenshots, product captures, b-roll references. The skill pulls them in and places them at the timestamps where they match the script beat.

Step three: captions. The aligned word timestamps generate the caption track automatically. No manual timing, no guessing where lines should break.

Step four: sound. The skill selects sound effects based on the script's pacing and picks a music bed from your configured library. The music comes in at the right beat and cuts where the script calls for silence.

Step five: render. The edit plan compiles to a HyperFrames scene. The render produces a 1080x1920 MP4 at 30 fps, mixed to -14 LUFS. The file lands in the output folder you configured.

Total time: roughly 2.5 seconds of render per second of finished video, plus the asset prep.

The Edit Plan

Before it renders, the skill produces edit.json. This is the full plan: every cut, every caption line, every sound cue, every music start and end. You can review it before the render runs. If one beat is wrong, fix the JSON and re-render that segment.

This is the quality lever. Review the plan, not the rendered output. Catching problems at the plan stage is faster than re-rendering.

What You Get Back

1. edit.json: the complete edit plan with every decision documented
2. A rendered 1080x1920 MP4 with captions, SFX, and music mixed in
3. A contact sheet showing every frame used, timestamped
4. A run log listing every asset used, every sound placed, and the music bed chosen
5. The finished file ready for upload, nothing else needed

How to Use It

1. Prepare the script as a plain text file.
2. Provide the voiced recording (MP4 or clean MP3).
3. List any link-based b-roll or screenshots you want used.
4. Run the skill, specifying the output Drive folder.
5. Review edit.json before render if you want to catch structural issues early.
6. Check the finished MP4 once. If a specific cut is off, adjust the timestamp in the edit plan and re-render that segment.

Done.

Save It as a Skill

After your first run, save the configuration: music library preferences, caption style, default b-roll sources. Every future run picks up those settings. You update the script and the recording. Everything else is already set.

The more you run it, the less you adjust. By the third video, the output matches your style without any manual correction.

Honest: What This Does Not Do

The skill executes what you give it. It does not restructure a rambling script. If the b-roll links are thin, the visual sequence will be thin. If the audio quality is poor, the word alignment degrades and the captions drift.

Clean inputs produce a clean output. Messy inputs produce a messy one.

The machine also does not make judgment calls about performance. If your delivery was flat on a particular line, the skill does not slow the cut to compensate. That is still a human call.

One Action This Week

If you have a recording sitting unedited in your Drive right now, run the skill on it. The API cost is cents. The time you save is the full editing session.

If you pay a human editor per video, this is the week to find out what that cost looks like at zero.

The skill package is linked at the bottom of this page. It includes the engine-neutral setup for Claude Code and Codex, the configuration file, and the first-run checklist.