NAREG TOKATLIAN

Automated AI Multimedia Pipeline

Svelte.js,  Bun.js,  TypeScript,  Tailwind CSS,  Postgres DB
video generator


Project Overview:
A full-stack video orchestration platform designed to automate the production of YouTube and TikTok content. This project transitions away from traditional request-response cycles by utilizing a Bun-powered WebSocket server, providing a persistent bi-directional link for real-time status updates during long-running AI generation tasks.


Technical Highlights:

Real-Time Orchestration: Engineered a custom WebSocket API using Bun to stream generation progress (image synthesis, audio encoding, and video rendering) directly to the client.

Reactive Frontend: Built a high-performance UI in Svelte that handles complex state management for multi-track video customization, including narration styling and subtitle synchronization.

Data Integrity: Implemented a PostgreSQL schema designed for high-concurrency, managing user configurations, asset metadata, and generation history.

Granular Control: Developed an intuitive interface for "human-in-the-loop" AI generation, allowing for precise overrides of image prompts, voice modulation, and subtitle timing.