Why sixthstump?

Cricket commentary is an art form. The way Ravi Shastri brings the drama, the way Nasser Hussain reads the game, the way Ian Smith captures a moment. It turns a sport into a story. I wanted to see if AI could capture even a fraction of that magic.

This started as a weekend experiment born from two things: a deep love for the game, and curiosity about where voice AI is headed. I wanted to explore how far text to speech has come, understand the APIs behind real time sports data, and build something that actually feels like sitting down and watching cricket with great commentators.

The backstory

I ran the Harvard Cricket Club in college, organizing matches, recruiting players, and doing whatever it took to keep cricket alive on a campus that didn't always know what a silly mid on was. That love for the game never went away. Now I work as a trader in New York, but cricket is still the thing I'll stay up until 4am for.

sixthstump is what happens when you combine that obsession with a fascination for what AI can do with voice, emotion, and real time data.

How it works

• ESPNcricinfo provides the raw ball by ball data, scraped in real time as each delivery happens
• Python / FastAPI powers the backend server, orchestrating the scraping, commentary generation, and audio streaming pipeline
• Claude API takes each delivery and generates multi voice commentary with emotion, context, and natural banter between commentators
• Cartesia TTS synthesizes speech with three distinct commentator voices (Ravi, Nasser, and Ian), each with unique tone and pacing
• WebSockets stream audio and text to the browser in real time so commentary plays in sync with the action
• Web Audio API handles playback on the client, queuing and decoding audio segments for seamless, gapless commentary