Multilingual Live AI Web Application
A React 19 web app designed to work in tandem with the JARVIS local AI cluster — centralizing everything a professional or learner needs to work in any language, in real time.
Features · Quick Start · Interface · Tech Stack · API · Ecosystem
Lumen bridges the gap between humans and languages. It captures your microphone input, transcribes it live with Whisper large-v3, translates it across languages in real-time, applies contextual grammar corrections, and speaks the result back — all in a single responsive interface.
Built for the JARVIS cluster, it leverages local AI inference for privacy and speed.
| Feature | Description | Tech |
|---|---|---|
| Live Transcription | Mic capture with WhisperFlow (Whisper large-v3) — low-latency streaming | Whisper + WebSocket |
| Simultaneous Translation | Multi-language real-time translation powered by JARVIS AI models | LLM Inference |
| Grammar Correction | Contextual AI-powered grammar and style suggestions | NLP Pipeline |
| Voice Synthesis | High-fidelity TTS via Piper — natural-sounding output | Piper TTS |
| Responsive UI | Modern React 19 interface — works on desktop, tablet, and mobile | React 19 + Tailwind |
| WebSocket Streaming | Real-time bidirectional communication for instant feedback | ws + Express |
| Docker Deployment | One-command deployment with docker-compose | Docker Compose |
| Language Detection | Automatic source language identification | Whisper built-in |
Lumen provides a clean single-page experience with real-time feedback.
┌──────────────────────────────────────────────────────────────┐
│ LUMEN — Multilingual Live AI │
│ │
│ ┌────────────────┐ ┌──────────────────────────────────┐ │
│ │ Microphone │ │ Transcription Feed │ │
│ │ [ Record ] │ │ "Bonjour, comment allez-vous?" │ │
│ │ ~~~~~~~~~~~~ │ │ > Hello, how are you? │ │
│ └────────────────┘ │ > Hola, como estas? │ │
│ └──────────────────────────────────┘ │
│ ┌──────────────────────────────────────────────────────┐ │
│ │ Grammar: ✓ No corrections needed │ │
│ │ Audio: [▶ Play translated output] │ │
│ └──────────────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────┐
│ Frontend │
│ ├── React 19 + Vite 5 (fast HMR) │
│ ├── TailwindCSS 3 (utility-first styling) │
│ └── WebSocket client (real-time streams) │
├─────────────────────────────────────────────────────┤
│ Backend │
│ ├── Node.js + Express │
│ ├── WebSocket server (bidirectional audio/text) │
│ └── REST API for config and session management │
├─────────────────────────────────────────────────────┤
│ AI Layer (JARVIS Cluster) │
│ ├── WhisperFlow — live transcription (large-v3) │
│ ├── Translation models — multi-language │
│ ├── Grammar models — contextual correction │
│ └── Piper TTS — high-fidelity voice synthesis │
├─────────────────────────────────────────────────────┤
│ Infrastructure │
│ ├── Docker + docker-compose │
│ └── JARVIS cluster integration │
└─────────────────────────────────────────────────────┘
git clone https://github.com/Turbo31150/lumen.git
cd lumen
docker-compose up -dOpen http://localhost:3000.
git clone https://github.com/Turbo31150/lumen.git
cd lumen
npm install
npm run devRequires a running JARVIS cluster for AI features (transcription, translation, TTS).
| Language | Code | Transcription | Translation | TTS |
|---|---|---|---|---|
| French | fr |
Yes | Yes | Yes |
| English | en |
Yes | Yes | Yes |
| Spanish | es |
Yes | Yes | Yes |
| German | de |
Yes | Yes | Yes |
| Italian | it |
Yes | Yes | Yes |
| Portuguese | pt |
Yes | Yes | Yes |
| Japanese | ja |
Yes | Yes | Yes |
| Chinese | zh |
Yes | Yes | Yes |
| Arabic | ar |
Yes | Yes | Yes |
| Russian | ru |
Yes | Yes | Yes |
| + 10 more | ... | Yes | Yes | Partial |
| Endpoint | Method | Description |
|---|---|---|
/api/transcribe |
POST |
Submit audio for transcription |
/api/translate |
POST |
Translate text between languages |
/api/grammar |
POST |
Grammar check and correction |
/api/tts |
POST |
Text-to-speech synthesis |
/api/languages |
GET |
List supported languages |
/api/health |
GET |
Service health check |
/ws |
WS |
Real-time bidirectional stream |
The Lumen interface provides a clean, single-page experience:
- Microphone panel — tap to record, see live waveform
- Transcription feed — real-time text as you speak
- Translation output — select target language, see instant results
- Grammar overlay — corrections highlighted inline
- Audio playback — listen to the translated/corrected output
Lumen is part of the JARVIS distributed AI cluster:
| Project | Description |
|---|---|
| jarvis-linux | Distributed Autonomous AI Cluster |
| TradeOracle | Autonomous Crypto Trading Agent |
| lumen | Multilingual Live AI Web App (this repo) |
| gemini-live-trading-agent | Voice Trading Assistant |
| gemini-creative-storyteller | Interactive AI Storyteller |
| browser-mcp-orchestrator | Dual-Browser DevTools Orchestration |
| transcription-multi-langue | Lightweight Multilingual Transcription |
MIT (c) 2026 Turbo31150 — Franck Delmas
Freelance profile: codeur.com/-6666zlkh