Automating My Product Demo Videos

I like building products more than I like rebuilding the same launch video by hand. This private tooling project turns a repeatable browser walkthrough into a composed product demo, so a new version of an app does not require starting the recording and layout process from zero.

The repetitive part after shipping

A usable demo needs more than a screen recording. The browser has to start from a known state, the viewport has to be consistent, interactions need to happen at the right time, and the final footage needs to fit the layout and pacing of the post.

Doing that manually once is fine. Doing it after every meaningful product change is exactly the kind of work I tend to automate.

I built the pipeline around repeatability. A scripted walkthrough opens the product, performs the important interactions, and captures the result with a controlled viewport. The rendering step then turns that capture into the final composition. Product-specific actions can change, but the framing and output workflow remain reusable.

Known app state
      ↓
Scripted browser walkthrough
      ↓
Viewport recording
      ↓
Composition and render
      ↓
Shareable demo clip

Why the source stays private

The pipeline is internal tooling rather than a packaged public library. It contains project-specific recording sequences and assumptions about how I launch and capture my own applications. The interesting portfolio artifact is the workflow and its output, not publishing configuration that would be useful only inside my setup.

This demo shows the pipeline being used for LARPmaxxing:

The hard part is timing

Browser automation is deterministic until time enters the picture. A page may be ready before a font, animation, or network-backed result has settled. Recording adds another clock, and rendering adds one more. If those clocks disagree, the output can look choppy even when every individual step “worked.”

That is the current engineering edge of the project. The workflow already removes the layout and interaction repetition; the next pass is improving frame pacing so the recording looks as consistent as the scripted behavior.

I like this project because it is tooling built in response to a real annoyance. It does not need to become a giant video platform. It only needs to make publishing the next good demo cheaper than putting it off.

Tech used

Node.js, browser automation, scripted capture, video rendering, and product-specific test flows.