ContextLens is a real-time conversation assistant that provides gentle, optional social-cue support by analyzing speech, body language, and environmental context. Built as a lightweight web app (demoed on Meta Quest 3), it helps users—especially introverts and neurodivergent individuals—navigate social interactions with more clarity and confidence.
- Real-time rolling transcript (few lines at a time)
- Subtle corner popups for social cues (only when needed)
- Speech analysis + transcription (Deepgram)
- Body language + environment understanding (Gemini Vision)
- Multimodal reasoning layer with large context window (Gemini)
- Optional live translation support
- Frontend: Web app (npm)
- Deployment: Vercel
- Speech Transcription: Deepgram
- Vision Understanding: Gemini Vision
- Reasoning / Brain: Gemini
- Audio Capture (optional): SoX
- Node.js 18+
- npm
- API keys:
- Deepgram
- Gemini