WhisperWriter

An open-source dictation app types speech into your active window using local Whisper models on CPU or NVIDIA GPUs, or through OpenAI's API.

WhisperWriter turns microphone speech into typed text in your active window. It's for people who want dictation in their desktop apps, with a choice of transcription on their own computer or through an API.

Local transcription: Run Whisper through faster-whisper on a CPU or NVIDIA GPU, choosing a model size to balance recognition accuracy against speed.
Recording controls: Use a customizable keyboard shortcut for continuous dictation, recording that stops after silence, toggle recording or hold-to-record input; continuous mode transcribes after a pause and resumes listening.
Recognition context: Specify a language and an initial prompt to guide transcription, with local options to use earlier text as context and filter out silence.
API transcription: Connect to a local endpoint such as LocalAI, or send recordings to OpenAI's cloud API, which requires an API key.

Local transcription is the default and processes audio on your computer. The app accepts a model stored locally; otherwise, it downloads the default model. NVIDIA GPU processing requires CUDA libraries. WhisperWriter is a Python app, is open source under GPL-3.0 and has a settings window plus a status window you can hide during use.