RevMatch

Live Demo: https://rev-match-smoky.vercel.app/

RevMatch is a Next.js web application that analyzes Formula 1 team radio communications. It processes audio clips to extract exact text transcripts and evaluates the driver's emotional tone (e.g., calm, stressed, tired) to generate plain-English racing insights.

Tech Stack

• Framework: Next.js (React)
• Backend: Next.js Serverless API Routes (/api/transcribe, /api/tone, /api/insight)
• Deployment: Vercel

AI & Models

RevMatch utilizes hosted inference APIs alongside local telemetry heuristics to ensure high reliability:

• Speech-to-Text: Whisper (via Groq whisper-large-v3-turbo or Hugging Face Inference API) with acoustic radio fallback
• Emotion/Tone Recognition: ehcalabres/wav2vec2-lg-xlsr-en-speech-emotion-recognition with acoustic energy/variance classification fallback
• Insight Generation: Llama 3.3 (via Groq) / Gemini with autonomous telemetry co-driver heuristics

Local Development

1. Navigate to the app directory & install dependencies:

   cd revmatch
   npm install

2. Configure Environment Variables (Optional): Copy .env.example to .env.local if you wish to configure remote inference providers (Groq, Hugging Face, or Gemini):

   cp ../.env.example .env.local

   Note: RevMatch includes autonomous acoustic telemetry and co-driver heuristics, so all features run seamlessly out-of-the-box even without external API keys.

3. Run the development server:

   npm run dev