screamer
Windows dictation that types where the cursor already is: hold a key, speak, release. Bring your own speech-to-text endpoint.
- Source code0 stars
Commercial dictation tools want a subscription and their own cloud. Screamer does the same job against any Whisper-compatible endpoint - Groq, OpenAI, or anything else that speaks the same protocol - so which provider you use, and what it costs, stays your decision.
Hold the hotkey, speak, release. It records 16 kHz mono audio, sends it for transcription, optionally runs an LLM pass for spelling and grammar, and types the result into whatever window has focus. It lives in the system tray, does push-to-talk or toggle, falls back to a second provider when the first one fails, and keeps API keys encrypted with Windows DPAPI. Written for my own use, MIT licensed.