Skip to content

Herk Edit Kit by Nate Herk: edit videos, reels, and YouTube Shorts with Codex or Claude Code. 15 skills, transcript-driven cuts, storytelling, 406 motion-graphics cards, and HyperFrames templates.

1.9 KB
Raw Blame

Setup and troubleshooting

Start with tools, accounts, and API keys for Nate's ElevenLabs and Kie.ai choices, Whisper alternatives, and optional services.

Use Node 22+ (required by the pinned HyperFrames CLI), FFmpeg with ffprobe, and Chrome/Chromium. Run npm ci in the repository root before any project commands. npm run setup checks tool availability; npx hyperframes doctor diagnoses the renderer. If Chromium is unavailable, follow npx hyperframes browser --help. Windows users should reopen their terminal after installing FFmpeg to refresh PATH.

The repository pins HyperFrames 0.7.109 and GSAP 3.14.2. Keep package-lock.json with package.json. Upgrade deliberately, then repeat the render checks.

npm run new-video -- lesson-one copies the synthetic starter and local GSAP into a new self-contained folder. Existing project names are refused so nothing is overwritten. Replace the starter with your recording and planned compositions. The left glass template needs your own assets/source.mp4 with an audio stream.

Codex reads AGENTS.md and discovers .agents/skills/. Trust the project in Codex if you want its project settings loaded. Restart if the skills do not appear. Claude Code reads CLAUDE.md and .claude/skills/. Neither runtime needs a special MCP server for local editing. The root npm run check validates the distribution; it does not lint personal work under video-projects.

To update a skill, edit .claude/skills/<name>/, run npm run sync:skills, and then npm run check:skills. No Windows symlink privileges are needed.

Scribe errors: a 401 generally means invalid authentication; a 403 can mean missing speech_to_text access. Check the key's permissions without printing it. No key is required for npm test, starter preview, or starter rendering. For speech-free sources, supply a separate audio track before using the cutting renderers: they expect one video stream and one audio stream.