Real-time accessibility guidance for the visually impaired — scene analysis, obstacle detection, OCR, and navigation — all powered by AI, available in 4 languages.
Globally, 285 million people live with visual impairment. Existing assistive devices cost $3,000–$5,000, putting them out of reach for most of the world.
SightlineAI transforms any camera-equipped device into a real-time accessibility companion — analyzing scenes, reading text, detecting obstacles, and guiding users with AI.
Camera captures the environment while hardware sensors (LiDAR/IMU) detect obstacles and terrain — all processed in real-time.
Qwen AI models analyze the scene for objects, text (OCR), hazards, and contextual understanding — with offline fallback for unreliable connectivity.
A dedicated risk engine evaluates danger levels, accessibility scores, and generates prioritized guidance — keeping safety first.
Contextual, multi-language voice and text guidance delivered via WebSocket streaming with conversation memory for natural interaction.
16 core features across AI, input processing, contextual awareness, and platform capabilities.
State-of-the-art Qwen language model for natural, context-aware scene understanding and conversation.
Real-time scene description, object detection, and spatial understanding from camera input.
Multi-turn dialogue with memory for natural follow-up questions about your environment.
Intelligent offline mode with cached responses and local processing when connectivity drops.
Real-time frame capture and preprocessing optimized for low-light and varied conditions.
Read signs, labels, documents, and screens with high-accuracy optical character recognition.
LiDAR and IMU sensor integration for precise obstacle detection and terrain analysis.
Hands-free voice commands with noise cancellation for natural interaction while navigating.
Real-time location context, nearby points of interest, and accessibility-aware routing.
Dynamic danger evaluation with severity scoring and prioritized safety guidance.
Evaluate routes and locations for accessibility with quantitative scoring metrics.
Step-by-step voice navigation with obstacle avoidance and terrain-aware routing.
Full support for English, Bangla, Arabic, and Spanish with automatic detection.
Real-time bidirectional streaming for instant AI responses without polling delays.
Secure user accounts with JWT authentication, session management, and data privacy.
Save locations, conversations, and preferences for personalized accessibility guidance.
End-to-end pipeline from user input to intelligent guidance response.
SightlineAI is fully open source under the MIT license. We believe accessibility tools should be transparent, community-driven, and freely available. Whether you're a developer, designer, or accessibility advocate — there's a place for you.
🤝 Contribute on GitHub