Home > Portfolio > TRANSCRIBELIVE: SPEECH TRANSLATION
AI Whisper WebRTC Node

TRANSCRIBELIVE: SPEECH TRANSLATION

Real-time voice-to-text translator running speech-to-text models on active audio streams.

TRANSCRIBELIVE: SPEECH TRANSLATION Hero Image

TranscribeLive takes audio feeds from WebRTC streams, feeding them to Whisper models for translation.

"Features sub-second latencies by processing audio inputs in overlapping 2-second buffer files."

WebRTC Pipeline

Combines React, Node.js, and Python. Audio streams are routed using WebRTC, sliced into raw buffer segments, and processed with faster-whisper pipelines.

System Architecture

[CLIENT AUDIO CAPTURING]
React.js with MediaRecorder API / WebRTC audio capture tracks
▼
[REAL-TIME STREAM WORKER]
Node.js WebRTC server / PyTorch Faster-Whisper pipeline / 2-second buffers
▼
[REAL-TIME TRANSCRIPT DELIVERY]
Socket.io Event Emitter (Direct screen text stream injection)