Show HN: Self-hostable STT service with a Deepgram-compatible listen API (vedgupta.in)

Whisper API, an open-source and self-hosted speech-to-text service, has been launched, enabling users to maintain full control over their audio and data while leveraging familiar integration patterns through a Deepgram-compatible API. The service utilizes whisper.cpp for its backend and offers real-time transcription capabilities over WebSockets, as well as rich output formats including JSON, SRT, and VTT. This compatibility with the widely used /v1/listen endpoint streamlines the integration process for developers, ensuring a seamless transition to an in-house solution. The significance of Whisper API lies in its flexibility and security features. It allows for advanced functionalities such as custom vocabulary prompting and audio windowing, alongside robust security measures like API key management and strict limits on URL ingest to mitigate risks. Built using modern tech stacks like FastAPI and SQLAlchemy, the Whisper API not only supports easy local deployment and testing but also provides comprehensive documentation for users, fostering further innovation in the AI/ML community. This initiative reflects a growing trend towards democratizing access to speech-to-text technology, empowering developers to create customized solutions while adhering to data privacy standards.