Ghostlens

screenshots with all the featurs avail, right (after redaction)

Inspiration

With the rise of AI, personal data is being collected, analyzed, and shared faster than ever, often without users’ awareness. We wanted to build a solution that gives people control over their privacy and protects them from accidental data exposure.

What it does

GhostLens is an AI-powered privacy protection tool that helps TikTok creators safely share videos without accidentally exposing sensitive information. It scans each video frame by frame to automatically detect private data such as faces, addresses, emails, phone numbers, IDs, and even spoken information through speech-to-text. With one click, creators can apply smart privacy presets that instantly blur, bleep, or mask sensitive content, making it effortless to secure videos without manual editing. GhostLens also calculates a real-time Privacy Score to show how safe a video is before posting and generates a downloadable privacy report with timestamped audit logs for full transparency. Finally, it exports a ready-to-post sanitized video, enabling creators to focus on storytelling while GhostLens handles privacy seamlessly in the background.

How we built it

Frontend: Built with React + Vite + TypeScript and styled using Tailwind CSS for a fast, clean, and responsive interface. It manages video playback, overlays (blur, bleep, mask), the detection timeline, and Privacy Score display. Backend: Powered by Python + FastAPI, responsible for handling video frames, processing detections, and returning timestamped results. AI & Detection Layer: Face & person detection to identify identities on-screen PII detection (emails, phone numbers, addresses, IDs) via custom NLP models + regex rules Whisper integration for speech-to-text, enabling detection of spoken PII Risk scoring logic to calculate a real-time Privacy Score Demo Setup: For the hackathon, GhostLens runs locally, the frontend and backend communicate via REST APIs, with no external hosting required.(can be adapted for future cloud deployment)

Reports & Export: Creates sanitized videos ready to post and a downloadable privacy report with timestamped audit logs.

Challenges we ran into

Training our AI model to accurately detect different formats of sensitive data. Ensuring real-time redaction without sacrificing speed or user experience. Debugging API integration issues between frontend and backend. Integrating multiple detection systems, combining face recognition, PII detection, and speech-to-text while keeping performance smooth.

Accomplishments that we're proud of

We built GhostLens, an AI-powered privacy tool that detects and protects personally identifiable information (PII) in videos with over 90% accuracy. We successfully integrated real-time video analysis, face and PII detection, speech-to-text processing, and one-click redaction into a seamless, creator-friendly platform. We also optimised video processing for smooth frame-by-frame analysis and designed a simple, intuitive interface that makes privacy effortless for creators.

What we learned

Through building GhostLens, we gained deep insights into integrating AI-powered privacy tools into a real-time video processing system. We learned how to balance performance, accuracy, and user experience while ensuring seamless integration of features like face detection, PII recognition, and instant redaction. This journey also reinforced the importance of data security, ethical AI design, and user-focused innovation.

What's next for Ghostlens

Ghostlens is evolving into a next-generation privacy platform. We’re launching a browser extension for instant, one-click redaction on any website, integrating AI-driven, user-controlled privacy policies that adapt intelligently to different contexts, and expanding towards a cross-platform ecosystem that delivers a seamless, unified privacy experience across web, mobile, and enterprise environments, positioning Ghostlens as the go-to privacy engine for creators, businesses, and everyday users.