A Video Production Pipeline Built Around a 15-Minute Constraint

Goal

Create educational videos and shorts consistently without spending hours editing.

The original motivation was supporting content distribution for ClearFit, a career-fit project I was experimenting with. I wanted a workflow I could realistically sustain long-term.

Target:

• less than 15 minutes of effort per video
• good AI voice quality
• good quality AI-generated imagery
• visually engaging output
• inexpensive to run locally (I was curious how much quality could be achieved with roughly 16GB of VRAM before reaching for paid APIs)

Approach

After testing several text-to-speech systems and image-generation workflows, I settled on a combination that produced acceptable quality on local hardware.

The central piece became a JSON specification describing the entire video.

Each file contains:

• title
• scenes
• video thumbnail
• narration
• image prompts
• animation instructions

System

The JSON acts as the source of truth.

From there, local scripts:

• generate narration
• generate images
• animate backgrounds
• assemble scenes
• render the final video

A single specification can produce both long-form videos and shorts.

Result

The workflow is now simple enough that a new video starts as a JSON file and can be rendered with a few commands.

More importantly, the production process fits the original constraint: creating content quickly, inexpensively, and with little ongoing effort.