themanyone/whisper_dictation

Leverages **whisper.cpp** and **llama.cpp** for offline speech-to-text and LLM inference without PyTorch dependencies, enabling concurrent AI workloads on resource-constrained hardware. Provides system-level voice control through `pyautogui` and `evdev` (with automatic X11/Wayland detection), integrating with external services like Stable Diffusion WebUI, mimic3 TTS, and optional cloud APIs. Uses GStreamer for audio capture/processing and exposes whisper via a local HTTP server for client communication.

289 stars.