Experiment

Teti Speech-to-Text Tool

Browser speech-to-text hobby tool for recording or uploading audio and video, then transcribing with Grok's whisper-large-v3 model using a bring-your-own API key.

Overview

Teti is a small browser-based transcription tool for turning recorded speech or uploaded audio/video files into text. It is designed around a bring-your-own Grok API key flow, so users can test transcription without a platform account system.

Technical Details

Built with Next.js, Tailwind CSS, and Grok's transcription API, the app supports:

Microphone recording directly in the browser
Audio and video file uploads for transcription
Grok whisper-large-v3 transcription through Next.js API routes
Copy-to-clipboard transcript handling
Local encrypted API-key storage using AES-GCM, IndexedDB, and browser storage

Product Notes

The important product constraint was trust: the API key flow needed to feel lightweight while avoiding plain-text local storage. The current implementation keeps the key client-owned, encrypts it locally, and routes transcription requests through the app's API layer.

Status

Personal hobby tool, live on Vercel and open on GitHub.