VoiceScript

Transcribing audio manually is slow and error-prone — whether it's a lecture, a meeting, or a voice note. VoiceScript was built to make that instant, with no installs and no cost to the end user.

The core flow is simple: record directly from your microphone or upload an audio file, and the app sends it to Groq's Whisper Large v3 Turbo model — one of the fastest and most accurate speech recognition models available — and returns a clean transcript in seconds.

Every transcription can be copied to clipboard, downloaded as a .txt file, or saved to a PostgreSQL database hosted on Neon. A dedicated History page lets users revisit, re-download, or delete any past transcription.

The frontend is built in React with a dark, animated UI — waveform visualisation during recording, smooth page transitions, and toast notifications for every action. The Flask backend handles audio processing and database operations, deployed on Render with a self-ping mechanism to prevent cold starts.