Free Desktop Dictation
The native app is free to download. Local inference is free to run, private, and has no usage meter. Cloud transcription and the hosted API are optional paid services.
1. Install the app
curl -fsSL https://dictatorflowstatic.dictatorflow.com/install.sh | sh
This downloads the binary and desktop integration only. It does not ask for a DictatorFlow account or create a cloud API key.
2. Or grab the binary directly
No login wall. Manual downloads need PATH and desktop integration set up separately.
3. Turn on free offline mode
Open the app, go to Settings, and choose Local. The app downloads the speech model for you and fills in the model path. Nothing to install by hand.
One click, then it is offline forever
Free, private, no usage meter, no account. Runs on NVIDIA CUDA, AMD ROCm, or plain CPU, whichever your machine has. Prefer a terminal? Run dictatorflow --download-model.
Keyboard shortcuts
Ctrl+Shift+DDictate into any app
Ctrl+Alt+JTransform selected text with a voice command
Optional cloud mode
Want zero model setup, hosted accuracy, sync, or API access? Create an account and add a paid cloud plan or usage credits.
CONFIGURE CLOUDAdvanced: manage model files yourself
Optional. Only useful if you want a custom path, a pinned revision, or an offline machine. The in-app download covers everyone else.
Parakeet · the default model
Same ONNX model the app downloads. Point Model Path at the resulting model.onnx.
pip install huggingface-hub huggingface-cli download istupakov/parakeet-ctc-0.6b-onnx \ --local-dir ~/.dictatorflow/models/parakeet-ctc-0.6b
VibeVoice BitNet · experimental
CPU-only Linux evaluation path. Free and offline, but our initial nine-clip test reached 60.45% average WER, so it is not the default.
curl -fsSL https://dictatorflowstatic.dictatorflow.com/install-vibevoice-asr.sh | bash
GPU acceleration is optional; supported builds fall back to CPU if the requested provider cannot initialize.