codeforequity-at/botium-speech-processing

What it solves

It provides a unified, developer-friendly API to access various free and open-source Speech-To-Text (STT) and Text-To-Speech (TTS) services, reducing the need to manage multiple disparate tools for audio processing.

How it works

The project acts as an opinionated software stack that wraps several audio tools into a single API. By default, it uses Kaldi for speech recognition, MaryTTS for speech synthesis, and SoX for audio file conversion. It can be deployed via Docker and supports both local open-source engines and integrations with cloud providers like Azure, Google, and Deepgram. It also includes a file system watcher for batch processing and a WebSocket endpoint for real-time audio streaming.

Who it’s for

Developers building voice-enabled chatbot services (such as IVR systems), creators synthesizing audio tracks for tutorials, and testers performing automated testing of voice services.

Highlights

• Unified API: Single interface for multiple STT and TTS engines.
• Hybrid Support: Works with local open-source tools (Kaldi, MaryTTS) or cloud APIs.
• Real-time Streaming: Supports streaming audio via WebSockets.
• Audio Utility: Integrated SoX for audio file conversion and artifact addition.
• Automated Processing: File system watcher for automatic transcription and synthesis.