ContentCreate
A local content factory: script, voice, music, captions, video and publishing on one laptop GPU.
What it does
- Turns a script or spec into voiced, captioned, music-backed vertical video.
- Includes a recipe engine that takes a dish and produces a five-scene cooking Short (keyframes, image-to-video clips, narration, music, captions).
- Exposes its actions through an admin UI and an MCP server so an assistant can drive it.
- Runs every GPU job through a serial queue so an 8 GB card never holds two models at once.
Why it exists
It was built as the acquisition flywheel for the owner's own channels and to prove that a useful content pipeline can run with no per-clip API fee for most steps.
Stack and status
Stack
- Windows laptop with an 8 GB RTX-class GPU
- Express backend with a serial job queue
- PowerShell and Python tools around image, voice, music and video models
- FFmpeg for assembly and loudness
- MCP server
Status: In use
In daily use for the owner's own channels. The recipe engine produced its first finished video in October 2026. Some steps (for example image-to-video clips) use a paid API, roughly fifty cents per clip.
What is public and what is private
Public
Nothing. Described here as a pattern.
Private
The code, model folders, brand specs and access keys. Private.
Honest limitations
- Hardware bound: one GPU job at a time, so it is slower than a cloud farm.
- Image models occasionally paint fake text or watermarks, so output is checked by eye before shipping.
- Single-machine setup, not packaged for other people to install.
Related projects
Content Studio AI
A multi-client workspace for running one-person content companies: plan, generate, edit, approve and publish.
See the project →Life OS
A personal operating system: Portfolio Tracker, Rolodex, planner, notes and an AI assistant that asks before it changes anything.
See the project →Talk to us
Thinking about a local, low-cost content pipeline for your brand? Describe what you publish today.