About Gestura
We believe communication
is a human right.
Gestura was built to close the gap between the deaf and hearing communities — using AI trained entirely from scratch, with no shortcuts.
Our mission
Making sign language accessible to everyone
Over 70 million people worldwide use sign language as their primary form of communication. Yet most people around them can't understand a single sign. That disconnect is the problem we set out to solve.
Gestura uses a custom-trained LSTM neural network — built on 87,000 hand gesture images — to recognize ASL letters in real time, directly from your webcam. No external AI services. No data sent to third parties. Just a model we trained ourselves.
The team
Built by engineers who care
Gestura started as a passion project with a simple goal — build something that actually matters. We spent months training the model, iterating on the UI, and testing with real users.
Every part of the stack was built from scratch — the dataset preprocessing pipeline, the LSTM architecture, the FastAPI inference server, and the React frontend. No shortcuts, no pre-built models.
Under the hood
The technology behind Gestura
PyTorch + LSTM
Custom LSTM neural network trained on 87,000 images. Runs on Apple Silicon via MPS.
MediaPipe
Extracts 21 hand landmarks per frame in real time. No raw image pixels — just coordinates.
FastAPI
Python backend serving predictions via REST API. Processes each webcam frame in under 200ms.
React + Next.js
Modern frontend with live webcam feed, landmark overlay, and real-time prediction display.
Docker + AWS
Containerized deployment with GitHub Actions CI/CD pipeline to AWS EC2.
Tailwind CSS
Clean, responsive design system built with utility-first CSS. Works on all screen sizes.
See it in action
Open Gestura and start signing — no account needed.
Try Gestura free →