Orateur
Orateur is a local-first voice assistant that connects Whisper transcription, Pocket TTS, and Ollama-backed speech-to-speech into a desktop workflow with optional MCP tools, global hotkeys, installers, and a visible status layer for day-to-day use.
What needed to work
Voice assistants are often cloud-dependent, opaque, or locked into one interaction mode. Orateur needed to keep speech workflows local, fast to trigger, and understandable while still supporting extensible AI tooling.
How the product answers it
The system combines Whisper via pywhispercpp for STT, Pocket TTS for speech output, and an STT to Ollama to TTS pipeline for speech-to-speech. Global shortcuts and a Quickshell or Tauri UI keep the assistant available without forcing users into a dedicated app window.
What shipped
An open-source Python desktop assistant with release installers, configurable models and shortcuts, optional GPU acceleration, systemd support, and MCP tool access for richer local workflows.