Script in. Film out.

Content Machine

Hand your AI agent a script, in Nepali, English or any language. It reads the whole story, fixes every face and place, and returns images and video clips that look like they came from one film.

Get started See how it works
  • Your Google account
  • Your computer
  • Your approval before any video
generated from the line above

How it works

Every four words become a frame.

This is the real opening of the Okhaldhunga story. Scroll, and watch each four-word window turn into its image.

  1. Reads the whole script
  2. Locks every face and place
  3. Draws each four-word window
  4. Checks every frame by eye
c01-1

Stills

Each image carries the words it came from.

Sherpa migrations, Guthi, Teej, the Malla court. Every frame was generated from the line printed under it.

The reel

Then the frames start to move.

Clips from real projects, each one directed from a single line of script. Video always waits for your yes.

What makes it different

01

Reads first, draws second

It translates line by line and works out who is in the story, where and when, then asks every question in one go. Nothing is invented: no extra people, no made-up dates.

02

One look, locked

Each character gets fixed features and colours. Each place gets a "this, and not that" description. Those words are pasted identically into every prompt, so faces don't drift.

03

Many hands, one film

Helper agents write frames in parallel and images generate side by side. Because the look is locked first, speed never costs consistency.

04

Nothing leaves your side

It runs on your computer, bills your own Google Cloud project and saves to your disk. Video never runs until you approve the exact clips and cost.

Get started

Two things, then one line.

  1. A Google Cloud project with billing on

    Create one at console.cloud.google.com. The installer switches on Vertex AI for you. Google bills you directly; images cost a few cents each.

  2. An AI agent

    Claude Code (recommended) or Codex. The installer connects it for you.

irm https://content.tarjun.com/install.ps1 | iex

Paste it into PowerShell (Windows) or Terminal (Mac). It opens your browser to sign in with Google, asks which Cloud project to use, and connects your agent. On Windows it installs Node.js, the Google Cloud CLI and ffmpeg if they're missing; you may need to open a new window and run it once more. On a Mac, install Node.js and the Google Cloud CLI first.

Then just ask

"Use content-machine to make a 9:16 image pack from this script: …"

"Turn frames c03 and c07 of that pack into 6-second silent clips."

Images start straight away. Video always stops first, shows you the clips and the estimated cost, and waits. Everything lands in ContentMachine/Outputs in your home folder.

Updating, and when something's off

Update: run the same install line again. Your settings and your Outputs folder are kept.

Check: in a terminal, cd ~/ContentMachine then npm run doctor -- --online. Every ✘ line says what to fix. Paste the output to Claude Code if you're stuck.