CaptionCraft AI

**CaptionCraft AI**, developed by **Team Code Craft**, is an AI-powered video captioning platform that generates contextual captions using a multimodal AI pipeline. Instead of relying only on speech, it combines visual and audio understanding to create captions that accurately reflect a video's content. Each uploaded video first undergoes validation before entering the pipeline. Keyframes are extracted to capture important scenes, while the audio track is processed separately. If speech is detected, it is transcribed using an AI speech recognition model. The extracted keyframes and transcript, when available, are used to generate a summary of the video, which is then transformed into natural, context-aware captions. Videos without speech are also supported by generating captions primarily from visual information. To improve performance, CaptionCraft AI uses a four-worker parallel architecture. Each worker independently processes one video through the complete pipeline, enabling multiple videos to be captioned simultaneously and significantly reducing processing time. The backend is provider-agnostic, allowing seamless switching between **Google Gemini**, **Fireworks AI**, and **locally hosted language models** without modifying the pipeline. This enables deployment based on performance, cost, privacy, or offline requirements. Built with **FastAPI** and **React + Vite**, the application follows a modular architecture that is easy to extend with multilingual support, subtitle exports (SRT/VTT), timestamps, speaker identification, additional caption styles, and platform integrations. By combining multimodal AI, configurable LLM backends, and parallel processing, CaptionCraft AI provides a fast, scalable, and flexible solution for automatically generating high-quality video captions.