As a young entrepreneur and developer, I've always been fascinated by AI's potential to solve everyday problems. One of my recent projects was developing a Chrome extension that uses AI to convert speech to text for transcription. The idea came from my own need for fast, accurate transcription: traditional tools were either too expensive or not accurate enough, so I set out to build something better with modern AI.
I chose OpenAI's Whisper API for speech-to-text because it offers strong accuracy in Danish and English, fast processing, and straightforward integration with web apps. On the engineering side, the extension uses Manifest V3 for current Chrome support, React for the UI, TypeScript for safety, and the Web Audio API for recording. The interface is deliberately simple: users control recording, see transcription in real time, can export in useful formats, and the whole thing feels native in the browser.
The hardest parts were audio quality and speed. Real-world microphones and background noise vary a lot, so I added preprocessing, noise reduction, and tuning for different input devices. For responsiveness I leaned on streaming audio, parallel work where it helped, and caching so repeat work did not feel sluggish.
Shipping it taught me a lot about wiring AI into a real web product, building a Chrome extension end to end, handling audio capture responsibly, and keeping the UX obvious for people who just want text out of speech. Looking ahead, I want to broaden language support, add richer editing, connect to platforms people already use, and explore offline-capable paths where it makes sense.
For me the project is a reminder that AI is most interesting when it solves an actual friction people hit every day. If you want to try it, the extension is in the Chrome Web Store.