ANIMY

A web platform that turns a text prompt into a rendered video. It is split into three services: a React frontend, an API backend, and a separate render backend that does the heavy lifting.

I wanted to build a real generative-AI product end to end, not just a demo. That meant dealing with long-running render jobs and keeping the interface responsive while they run.

The render backend processes jobs across four parallel threads, which pushed throughput to roughly ten videos per hour. I evaluated two real-time update strategies for streaming progress to the client and dropped WebSockets after weighing the architectural cost against a simpler approach.

A working three-service pipeline with measurably higher render throughput, and a sharper instinct for when added complexity is not worth it.