Real-Time Sign Language Translation via WebRTC & Vision Models – Roast Uvilox AI!

Hey founders! I'm building Uvilox AI, an AI accessibility platform that translates sign language (hand gestures via MediaPipe/TensorFlow) into spoken voice and text in real time during live WebRTC calls.

The Concept:

• Problem: Over 430M people with hearing loss face major barriers in telehealth, emergency calls, and standard voice communication.
• Solution: A sub-second gesture-to-speech translation layer embedded directly into browser/mobile video calls.

Where I want brutally honest feedback:

1. Technical Viability: Is sub-second computer vision inference over live video streams scalable without massive GPU costs?
2. Product-Market Fit: Would emergency services or telehealth providers integrate an API for this, or is direct-to-consumer mobile app the better route?
3. User Experience: What critical edge cases in real-time gesture recognition am I likely overlooking?

Tear the idea, architecture, or business model apart!