# Codex and HyperFrames: our long-form editing workflow

From the Business Boomer video editing guide. Original workflow documented September 16, 2026. Commands and service availability can change; check the linked project documentation before installation.

## Long-form video editing with Codex: our HyperFrames workflow

For a recorded lesson, founder interview or screen-led tutorial, we now use **Codex to coordinate the edit in a local HyperFrames project**. You describe the result you want; Codex works with the footage, transcript, cutting scripts and editable motion graphics. HyperFrames renders the composition, GSAP animates its visual elements, and FFmpeg handles supporting media operations.

The starting point is **Nate Herk’s HyperFrames Student Kit**. Our setup combines that kit with ElevenLabs Scribe for transcription and optional Higgsfield imagery. The useful change is that editorial instructions, cut decisions and visual revisions can stay in one project instead of being reconstructed for each new export.

**What we have actually done:** our internal project produced an approximately 9-minute-40-second review export at 1280 × 720, with timed graphics and preserved approved audio. Its verification record includes spoken-anchor checks, video decoding and selected frame inspection. That establishes a working production workflow; it does not mean every frame has had a final editorial sign-off, or that an entire production finishes in one prompt. We still inspect the picture, listen to the joins and revise.

![Five stages of a long-form edit: source recording, transcript, reviewed cuts, editable graphics and final playback](/blog-images/video-editing-codex-workflow-20260916.webp)

### Start with Nate Herk’s tutorial and resources

**[Watch Nate Herk’s full video-editing tutorial on YouTube](https://www.youtube.com/watch?v=o3IEkKXXXvo)**, then keep the kit open alongside your own project. The video introduces the approach; the repository’s current instructions are the reference for setup and commands.

- **[Get the HyperFrames Student Kit](https://github.com/nateherkai/hyperframes-student-kit)** — the reusable editing skills, templates and example projects.
- **[Follow the installation guide](https://github.com/nateherkai/hyperframes-student-kit/blob/main/docs/SETUP.md)** and **[choose your tools and API services](https://github.com/nateherkai/hyperframes-student-kit/blob/main/docs/TOOLS-AND-API-KEYS.md)**.
- **[Work through the complete editing process](https://github.com/nateherkai/hyperframes-student-kit/blob/main/docs/WORKFLOW.md)** — including transcript timing and draft review.
- **[See the public showcase](https://github.com/nateherkai/hyperframes-student-kit/blob/main/examples/showcase/README.md)** for Nate’s short-form examples; these are separate from our internal long-form edit.

The kit supplies reusable building blocks. Choose the ones that explain your subject, and test them with your footage and text.

### Set up the same foundation

Install Node.js 22 or newer, Git, FFmpeg with ffprobe, and Chrome or Chromium. Then run:

```sh
git clone https://github.com/nateherkai/hyperframes-student-kit.git
cd hyperframes-student-kit
npm ci
npm run setup
npm test
npm run new-video -- my-video
```

Open the repository folder as a project in Codex. Place a copy of your recording at `video-projects/my-video/assets/raw.mp4`; keep the original unchanged. The Codex skills live in `.agents/skills/`, and `$edit-video` is the starting workflow.

For a first installation check, use `npm run demo` and follow the README’s preview and render steps. That starter does not require paid transcription or generated assets. Codex access and any external services have their own usage terms and costs.

### Give Codex an editorial brief

Copy this prompt and replace the path and subject:

```text
Use $edit-video for video-projects/my-video/assets/raw.mp4.
Create a landscape long-form edit about [subject] for [audience].
Preserve my meaning, useful examples and original recording.
Inspect the source and reuse a verified word-level transcript if available.
Otherwise propose ElevenLabs Scribe transcription and its cost first.
Trim dead air conservatively and show me mistake cuts in context.
Use the retimed transcript after each cutting stage.
Plan graphics around exact spoken phrases. Keep faces and demos visible.
Use a consistent visual style, readable text and purposeful motion.
Use existing assets first; propose any Higgsfield assets and credit cost.
Preview, render a draft, check frames and audio joins, then revise.
Keep a verification note and return the MP4 plus editable project.
```

The specific direction matters. Explain the intended audience, passages that must stay, visual style and target format before asking for effects.

### The editing sequence we use

| Stage | What Codex coordinates | What to check |
| --- | --- | --- |
| Understand the recording | Inspect media and obtain word-level timestamps | Correct names and technical terms; preserve the original. |
| Tighten the speech | Remove excess silence, then propose mistakes or retakes for review | Keep intentional pauses and repetition that help the explanation. |
| Rebuild timing | Align the transcript with the latest cut | Never use the original timestamps against shortened footage. |
| Plan the story | Map key ideas to spoken phrases and visual beats | Each visual should explain something rather than cover the speaker. |
| Build graphics | Assemble editable HyperFrames scenes and GSAP animation | Check type size, contrast, face placement and screen-recording detail. |
| Review and export | Preview, render a draft and inspect the encoded result | Listen for abrupt joins and check transitions, clipping and audio/video sync. |

For our workflow, **Higgsfield supplies optional supporting imagery; exact lettering and layout animate locally**. That lets us revise wording and timing without regenerating an entire scene. Nate’s kit also describes Kie.ai as an optional generation service. You do not need both providers to start.

Our selected transcription service is ElevenLabs Scribe. Configure your own credentials using the tools guide, keep them out of chat and version control, and approve media uploads and paid calls intentionally. An existing suitable transcript can avoid a new transcription request.

### Preview before you call it finished

From the repository root, check the project:

```sh
node scripts/preflight.mjs video-projects/my-video
cd video-projects/my-video
npx hyperframes lint
npx hyperframes preview
```

Review the sequence in Studio. Stop the preview when finished, then render a draft:

```sh
npx hyperframes render --quality draft --output renders/draft.mp4
```

For projects with anchored external sub-compositions, the workflow guide also documents `validate-beat-sync.mjs`. Inline compositions need manual timing review. A clean validator result cannot tell you whether a joke lands, an explanation makes sense or an audio cut sounds natural.

Watch the export and listen through the joins before final delivery. Save what passed and what still needs work in `VERIFY.md`. Once the long-form master is approved, use the kit’s **[short-form workflow](https://github.com/nateherkai/hyperframes-student-kit/blob/main/docs/SHORT-FORM.md)** to prepare separate vertical excerpts, with new caption and crop checks.

**Best fit:** teams comfortable working in a local project who want repeatable talking-head, tutorial and motion-graphics edits. Traditional editors remain useful for hands-on timeline work, advanced grading or specialist finishing. Start with one recording and judge the exported result before changing your whole production process.