How to clone a viral video with a coding agent

From a reference video to a rendered edit in one session: pick the reference, write the script, declare media, run the build, review the output.

Who this is for

You have a reference video whose format works and you want that format applied to your own product, presenter and audience. If you would rather prompt for footage from scratch, this is not the workflow you want.

What you need before you start

A reference video (your own or one you have the right to learn structure from), a written brief for the new video, and a coding agent that can run shell commands — Claude Code, Codex, Cursor or similar. Node.js 20 or newer.

Step by step

  1. 1. Install the skill

    One command registers the Hypit skill with your agent: npx skills add hypit-ai/hypit -g. The -g flag installs it globally so every project can use it.

  2. 2. Point the agent at the reference

    Tell your agent which file is the reference and what the new video is about. The reference supplies the beats; the brief supplies the content.

  3. 3. Write the script

    The script is the source of truth: segments for each beat, one or more takes per segment, role cues for speakers, and selections you can anchor tracks to later.

  4. 4. Declare media and generation

    Add the assets you already have as media items, and declare what should be generated. This is where you choose models — including your own provider keys.

  5. 5. Run the build

    A run produces the composition and renders it. Keep --keep when you want the intermediate results for review.

  6. 6. Review and refine

    Open the composition in Hypit Studio, adjust timing and appearance, and re-run. Because edits are anchored to words, changes stay local.

After this

The first clone is the expensive one. After that you change the script and re-run, which is why the workflow pays off around the third variant rather than the first.

Keep reading

FAQ

Does cloning a video copy someone else's content?

It copies structure — beats, pacing, caption rhythm — and fills it with your script and assets. Reproducing someone's footage or script verbatim is a copyright question, not a feature.

Do I have to pay to use this?

No. The agent tooling is free and open source; you can bring your own model keys. The managed API is optional.

How long does the first run take?

It depends on your models and how many takes you generate. The structural work — script, timing, tracks — happens before any paid generation.