The studio runs an internal pipeline that turns a one-line brief into a finished Instagram Reel, a multi-slide carousel and a caption — without manual editing. The same pipeline powers AptixLabs' own social presence and is the basis for the studio's social-automation offering for partners.
The pipeline, end to end
- A scheduled Cloud Run worker picks a brief from a Firestore queue
- An LLM drafts the full script and breaks it into shot beats
- Each shot pulls from a stock-clip library or generates b-roll on demand
- ElevenLabs synthesises voiceover in the studio's cloned brand voice, timed against the script
- Slides for carousels are composed in a headless browser, rendering templated HTML to PNG so typography stays on brand
- Output drops into Cloud Storage; the worker posts via the official Instagram Graph API
Why headless browser, not a video editor SDK
Headless Chrome renders templated HTML pages to PNG with pixel-perfect typography. A bespoke design template inside Figma would be hard to drive programmatically; HTML + CSS is. The same typography system the marketing site uses ends up in every carousel slide.
What it costs to run
About four cents in compute and AI calls per asset, including the ElevenLabs voiceover. Distribution via Instagram Graph API is free. The pipeline runs on a Cloud Run instance that idles at near-zero cost — only spinning up when a brief lands.
Where humans stay in the loop
A human reviews the daily queue once a week and can pull, edit or replace any asset before it ships. The point isn't to remove judgment, it's to make execution near-free so the team spends their time on judgment.