Skip to main content
karan.dev

Case 08In the open

← All work

SignSpeak

A polished chrome hand rising into frame, forming a two-finger sign, lit orange
00:00:00:00Play
Reel — SignSpeak08 / 08
Scope
Open-source build
Core stack
Python · MediaPipe · React
Role
Architecture, build, deploy
Links
Source ↗ github.com/karanxa1/SignSpeak (opens in a new tab)

The outcome

Real-time sign language translator: webcam hand tracking and a machine-learning pipeline convert American Sign Language gestures into English text and translated Hindi audio, delivered through a fast Vite frontend with a Python inference backend.

The build

Real-time sign language translator: webcam hand tracking and a machine-learning pipeline convert American Sign Language gestures into English text and translated Hindi audio, delivered through a fast Vite frontend with a Python inference backend.

Blueprint — how SignSpeak fits together

6 parts · 5 connections

Browser

Track

Infer

Output

Hands in, English text and Hindi audio out.

Hover, focus or tap a part of SignSpeak to trace its connections.

  • Vite frontend connects to Webcam
  • Webcam connects to MediaPipe
  • MediaPipe connects to ML pipeline
  • ML pipeline connects to English text
  • ML pipeline connects to Hindi audio

Credits — the stack

  1. Python01
  2. MediaPipe02
  3. React03
  4. Vite04
  5. Machine Learning05
  6. Computer Vision06
A polished chrome hand rising into frame, forming a two-finger sign, lit orange
Still — key artSignSpeak
Dozens of dark cables carrying small orange pulses, converging on a single bright ember point

Up next — 01

CallMissed.com