portfolio / ai video pipeline Lab
Remodelador: short video at scale, with human review
A pipeline that starts from a reference video, rewrites the script, generates image and voice, checks each step and asks for approval before publishing.

Numbers
- 1,297commits of mine since Sep 26, 2026
- ~1,200test files
- 56database migrations
The problem
Producing short video with AI goes through many paid steps: script, image, voice and editing. Done by hand, each video costs hours and each mistake costs money.
What I did
- The request comes in through an API, and a Python executor runs the steps one by one.
- A model checks the generated image before the pipeline moves on.
- A person approves the video before it is published.
- Production and staging are separate environments, with tests running on every change.
Result
The project is a partnership and does not generate revenue yet. It is my lab: what I learn there I apply to the clients’ agents. The code is private.
What I learned
Never repeat a paid call without knowing whether the previous one arrived. Each step records what it asked for before asking again.
Stack
- Cloudflare Workers
- Hono
- D1
- R2
- React
- Python