Tala: Free Audio-to-Text for My Designer Friend (No Signup, No Credit Card)

Developer Vicente Reyes has launched Tala, an open-source, privacy-focused audio-to-text tool designed to eliminate the friction of traditional transcription services. Built as a solution for a friend who required a simple, no-signup, and no-credit-card transcription platform, Tala utilizes OpenAI's Whisper model via faster-whisper to provide high-quality speech recognition. The application runs on a CPU-optimized backend using FastAPI and PostgreSQL, ensuring that audio files are processed locally and deleted immediately after transcription to maintain user privacy. Tala supports over 100 languages, offers subtitle file exports, and includes features like timestamp toggles and optional account creation for cross-device history. By leveraging open-source AI, the project avoids the costs and data retention policies associated with closed-source APIs, offering a transparent, self-hosted alternative for users needing quick, secure transcriptions. The full source code is available on GitHub, and the service is deployed via Render.