Boluto

Project overview

Boluto is a comprehensive, desktop-first AI video automation tool built to streamline the content creation workflow. It solves the time-consuming process of finding and editing highlights from podcasts, interviews, and lectures by intelligently transforming long-form media into viral-ready short clips (such as for TikTok, Reels, and YouTube Shorts). A core technical achievement of Boluto is its fully offline architecture; it operates entirely independently of cloud APIs, leveraging local hardware acceleration to ensure maximum data privacy and zero subscription costs. Under the hood, Boluto is powered by a sophisticated local backend service. The automated pipeline begins with audio extraction and high-accuracy transcription using whisper.cpp. It then employs a custom 4-pass LLM analysis framework (via Ollama) to deeply understand the video context, score chunk engagement, dynamically determine optimal clip boundaries, and rerank the outputs for quality and diversity. The final video processing is handled by FFmpeg, augmented with dynamic karaoke-style subtitles and optional MediaPipe/OpenCV face-tracking to automatically crop landscape videos into smooth, active-speaker-focused 9:16 portraits. The frontend is built to emulate the sophisticated aesthetics of high-end, professional video editing software. The UI features a robust workspace complete with a real-time Preview Console, A/B split viewing, a detailed Clip Strip, and comprehensive session management backed by SQLite. This dual-layered architecture—a heavy AI/video processing Python backend seamlessly bridged with a responsive modern desktop UI—demonstrates a deep technical capability in both low-level system automation and professional-grade user experience design.