How the Factory Works: A video studio that runs on one desk

The problem: Per-clip pricing adds up fast

Per-clip pricing adds up fast for a video studio running on a single desk. Every second of voice and each clip incurs a cost, making daily operations expensive. The AI-run factory has logged 12 short formats, each with unique elements like voice-overs and kinetic text. [1] [2] [3]

Each format includes six fast beats, each with a voice-over line and kinetic on-screen text. Five formats also require a generated image based on a shot prompt. This complexity multiplies the per-clip costs quickly. [3]

How the factory handles it

The factory handles per-clip pricing by running 215 finished shorts across 12 formats on a single machine. Each clip's production is logged as 'script local 9B model', indicating the AI script used. This streamlined process ensures efficiency and cost-effectiveness. [1]

For voice work, the log shows 'voice Kokoro, on device', signifying AI voice synthesis. Video production is logged as 'video Remotion, React', detailing the tools used. The system also includes 'checks plain code gates' to ensure safety and quality, maintaining a balance between automation and human oversight. [1]

Steal this: build it yourself

1. Set up local model. Install a local AI model for scriptwriting. This avoids per-clip costs. [1]
2. Run local voice. Use your machine to generate voiceovers, reducing per-second charges. [1]
3. Code-drawn video. Automate video creation with code, drawing slides, captions, and counters. [1]
4. Optimize formats. Create multiple video formats from the same code, maximizing efficiency. [1]
5. Monitor costs. Keep track of expenses to ensure per-clip pricing doesn't add up too fast. [1]