How the film was made

One message became a film in my voice

By Nicola Harvey

My graduation video was not really made by me sitting down to make it. I handed Joe's request to the operating system I have built over these months, what I call my BrainOS, and it came back as a finished film in my own voice. I barely steered it. Here is what actually did that.

The systems I built

My operating layer

A persistent identity and standing rules that load every session. This is what makes the work mine and not generic: my voice, what to keep private, how to route a task.

My second brain

Sixteen hundred conversations distilled into pages about me, my work and my history. This is where it pulled my whole body of work and my voice from, with nothing fed in.

My skill registry and model router

It picks the right tool and the right AI brain for each step on its own, and keeps private work on my own machine.

My subagent crew

My own delegation team, led by a Fable orchestrator. It ran several agents at once to gather everything, then handed the heavy build to the right specialist. That is how a big job got done in one sitting.

My voice clone lab

A local, private voice-cloning studio. It turned one short recording into my narrating voice, all on my machine, nothing uploaded.

My video toolchain

Films made entirely in code: title cards, motion, captions, transitions and rendering, all from a written brief.

My brand as code

My look is a design system, so every frame came out on-brand on its own.

Mission Control and a self-review pass

These caught every loose end, held the line on anything private, and reviewed the work honestly before it was called done.

The method
  1. One message in, read against my standing rules and my second brain.
  2. Fan out. Several agents at once gathered everything: my portfolio, every project on my machine, my systems, the graduation notes.
  3. Synthesise. All of it pulled into one clear inventory.
  4. My voice. Cloned my voice locally, then wrote the narration in my own words using my voice rules.
  5. Build. Handed the composition to my orchestrator, which built and rendered the whole film in code, on brand.
  6. Check. Reviewed every section frame by frame, held the privacy line, and logged it all to Mission Control.

The video is not the achievement. The real one is that I have built a second brain and an operating system that already know me well enough to take a single message and turn it into a finished, on-brand film in my own voice, and check its own work, with me barely in the loop. That is the whole point of what we have been building. The film is just it proving itself.