Hold a key, speak, release — clean, punctuated text lands right where your cursor is, in any app. Tap once more to send it to your AI. Runs entirely on your machine.
Bind any key or mouse thumb button. Hold it down — VocalCode starts listening instantly, no window in the way.
Say the code, the comment, the prompt. Release the key and clean, punctuated text is inserted at your cursor.
One more tap fires Enter — your words go straight to Claude, Cursor, Copilot, or whatever you're driving.
No subscription, no per-word metering, no server round-trip. The speech engine runs on your CPU.
$5, yours for good on this device. Not $10/month forever for something that should just be a tool.
Sub-200ms from release to inserted text on a normal laptop. Your voice never leaves the machine.
Speaks your language — Mandarin routed to Paraformer, English & European to NVIDIA Parakeet, per utterance.
Any text field in any app — your editor, terminal, browser, chat. It types where your cursor is.
Fixes spelled-out acronyms (“m c p” → MCP), adds punctuation and casing, and the talk→send loop fits AI tools.
No subscription. Free 30-day trial first.
🔒 Your voice never leaves this computer.
No. Speech recognition runs entirely on your computer. Nothing is uploaded — no audio, no transcripts.
A thin 6.9 MB installer. On first run it detects your machine and downloads the right speech model once — no bloat you don't need.
Chinese (Simplified) via Paraformer, and English plus 24 European languages via NVIDIA Parakeet. It auto-routes per utterance, or you can pin a language.
Any keyboard key or a mouse thumb button (X1/X2). Hold to talk; an optional second trigger taps Enter to send.
Yes — $5, once, for this device. Free updates within the major version. A future v2 would be a separate purchase, never a subscription.
Windows first. macOS is on the roadmap.