nateherkai/herk-edit-kit

Nate Herk's video-editing kit for Codex and Claude Code, built on HyperFrames. Formerly the HyperFrames Student Kit. Bring your own footage. Cut dead air, review mistakes, plan the story, and build motion graphics with HyperFrames and GSAP.

Four finished videos made with this kit.

Made with the motion-showreel skill: 15 seconds, 1920x1080, 60fps, cut to a 129 BPM grid. The red scrubber dot travels through seven craft chapters and lands as the play button in the logo.

The three short-form videos below were recorded by Nate. The first two were edited end to end by an AI agent running the editing skills in this kit, from his raw recording: cuts, captions, B-roll, motion graphics, music, and sound effects. Press Play on any video to watch it here in the README.

One take of Nate talking, edited by the agent into a 51-second reel about how it was edited.

Opens on the untouched raw recording, then hard-cuts to the finished edit and walks through what the agent did.

• Editing playbook: how Nate's videos get edited with AI, the measured house reel style, the B-roll, caption, and sound rules, and how every note becomes a rule in a skill.
• 15 skills, mirrored for both assistants, with their helper scripts and references.
• 406 draft motion-graphics cards across two styles, with manifests, CSS tokens, and editable slots.
• Two scene templates: dark graph paper and a left glass popout.
• Transcription, silence cutting, mistake detection, reviewed cut rendering, transcript retiming, EDL review, beat-sync validation, and preflight tools.
• Short-form editing: reels, YouTube Shorts, hook and payoff planning, precise captions, moving B-roll, and audio review.
• 12 existing teaching projects preserved from the original kit.
• A synthetic starter composition and editing fixture that need no footage or API key.

Nate uses ElevenLabs Scribe to transcribe and Kie.ai to generate videos and image assets. Bring your own API keys and credits when using those services. You can ask the assistant to use OpenAI Whisper or local Whisper instead, or supply an existing word-level transcript. Generated assets are optional.

The local starter needs no paid transcription or generation API. See the tools, accounts, and API-key guide for required tools, optional services, setup details, and prompts you can copy.

Install Node.js 22 or newer, Git, FFmpeg (including ffprobe), and Chrome or Chromium. Make node, ffmpeg, and ffprobe available in your terminal. Then run these commands in PowerShell, macOS Terminal, or a Linux shell:

git clone https://github.com/nateherkai/herk-edit-kit.git
cd herk-edit-kit
npm ci
npm run setup
npm test

Setup checks the tools and creates .env only if it is absent. Add only the keys for services you choose. The included transcription helper uses ElevenLabs; Whisper and Kie.ai integrations need the setup described in the tools guide. Setup and troubleshooting.

npm run demo
cd video-projects/demo
npx hyperframes lint
npx hyperframes preview

Scrub the eight-second animation in Studio. After reviewing it, stop the preview with Ctrl+C, then render:

npx hyperframes render --quality draft --output renders/demo.mp4

The demo uses local GSAP. HyperFrames may download and cache its font substitutions on the first render. It has no voiceover. The separate synthetic transcript exercises the cutting tools; it is fictional test data, not a transcript of the title animation.

Open this repository folder in Codex or Claude Code and say:

Use edit-video to edit my recording at [local path]. Keep my examples and core lessons. Tighten dead air, show me the proposed mistake cuts, and use the dark graph-paper style with occasional glass cards. Produce a reviewed draft.

Codex: $edit-video. Claude Code: /edit-video. For a single operation use cut-silences, cut-mistakes, video-storytelling, or style-library. See the step-by-step workflow, prompt recipes, and storytelling workbook.

Use short-form-edit to turn my recording at [local path] into a 9:16 reel. Build a truthful hook and payoff, preserve my meaning, tighten mistakes, and add precise captions, purposeful moving footage, and sound design. Prepare a draft for review. Use my existing footage before proposing generated assets.

Codex: $short-form-edit. Claude Code: /short-form-edit. The skill includes planning references and validators for caption timing, source mapping, scene coverage, and footage reuse. It is an agent-guided workflow; review the actual motion and audio before publishing. Short-form walkthrough and validation commands.

Use motion-showreel to make a 15-second showreel for [brand]. Study my reference reel at [local path] if I give one. Pick one motif that transforms through every chapter, cut on the music's beat grid, and end on the logo. Show me the storyboard and beat sheet before building, and confirm before any paid music or asset generation.

Codex: $motion-showreel. Claude Code: /motion-showreel. The skill includes a measured breakdown of the reel it was built from, a chapter technique library, a HUD template, and Node tools to analyze a reference video, measure a music take's beat grid, splice it onto the cut grid, and premix sound effects.

This is the main Herk Edit Kit repository. The newer video-pipeline kit has been merged here with both Git histories preserved. The original 12 projects remain in video-projects/, alongside the original shared brand examples and the make-a-video, short-form-video, and website-to-hyperframes skills. Use short-form-edit for new reels; short-form-video documents the older May Shorts compositions. Migration and compatibility notes.

The cards are reusable draft assets. Test the cards you choose with your text and footage. Library templates may load GSAP and Google Fonts from their public CDNs; localize these dependencies when assembling a final project.

The three showcase videos were explicitly supplied for public sharing. No other private recordings, transcripts, credentials, or private workspace settings were imported. Previously public examples remain in the repository. New folders under video-projects/ and raw-media/ are ignored automatically; files already tracked by Git remain tracked. Create a new project for your own footage.