DescX

DescX is an AI-powered video captioning platform built for creators, educators, developers, and accessibility advocates. Users upload a short video (30 seconds to 2 minutes), and the system automatically analyzes the visual content before generating four distinct caption styles: Formal, Sarcastic, Humorous (Tech), and Humorous (Non-Tech). The application follows a two-stage AI pipeline. First, a multimodal vision model extracts an accurate scene understanding from sampled video frames. That structured description is then passed to a dedicated language model that produces high-quality captions in multiple writing styles. This separation improves consistency, reduces hallucinations, and makes the system easier to extend with additional styles in the future. DescX includes an integrated caption editor, TXT export, and automatic burned-caption video generation through an FFmpeg-based rendering pipeline. The frontend is built with Next.js, while the backend uses FastAPI with a modular provider architecture that supports multimodal inference through Fireworks AI. The engine performs video validation, intelligent frame sampling, prompt construction, caption generation, formatting, and export in a complete end-to-end workflow. The project was designed with accessibility, simplicity, and extensibility in mind. Rather than generating only generic subtitles, DescribeX allows creators to instantly adapt the tone of their content for different audiences while maintaining a streamlined user experience suitable for both technical and non-technical users.