MojoFlow · Windows and Linux · Free
Hold a key. Speak. The words land where your cursor is.
MojoFlow is push-to-talk dictation that runs entirely on your machine. A speech model on your graphics chip or neural processor transcribes, a small on-device language model removes the fillers and applies your spoken corrections, and the text is typed into whatever app has focus. Nothing is uploaded.
A signed installer for Windows and a one-line install for Linux, both with in-app updates.
What it does
Dictation that respects the machine it runs on.
Push-to-talk or hands-free
Hold Right Ctrl, or any key you choose, and release. Hands-free mode taps to start and taps to stop. A small pill shows what has been heard so far.
Nothing leaves your PC
Speech and clean-up models run locally through Vulkan on any current AMD, NVIDIA or Intel graphics driver, with a CPU fallback. No account, no cloud, no upload.
A clean-up pass
Fillers go, punctuation arrives, and 'correct that, it's Thursday not Friday' replaces the last thing you said in place. Per-app rules set the style, or switch it off in a terminal.
Your words, your spelling
Teach it names and jargon under Dictionary, or from History with 'Teach a fix'. Spelling follows your Windows regional language: Australian, British, American, Canadian or New Zealand English.
Battery saver on Ryzen AI
On a Ryzen AI PC the same model can run on the neural processor through Windows ML, leaving the graphics chip idle and drawing a fraction of the power.
Dictate into AI Workbench
With AI Workbench on the same machine, hold the key with the Chat or Team Work composer focused and the cleaned-up text lands in the composer, attributed as dictation in the audit log.
Requirements
What it needs, and what it does not.
| Windows | Windows 10 or 11, x64, with the Edge WebView2 runtime (preinstalled on Windows 11) |
|---|---|
| Linux | A one-line install from the customer portal |
| Graphics | A driver with Vulkan for the clean-up pass; speech falls back to CPU without it |
| Models | Downloaded on first use: speech about 1.6 GB, clean-up 0.8 to 2.5 GB, chosen to fit your hardware |
| Not needed | No .NET, Python or Node install. No account. No internet after the models are down. |
| Price | Free |
Free, signed, and yours.
Install it, pick a key, say something. The first run walks you through the speech model, the microphone and a practice box.