Voice
Voice Mode
Dictate straight into the composer instead of typing. Hold the mic button to talk, or bind a toggle hotkey and use it hands-free — transcription runs entirely on your machine.
Hold to talk
Press and hold the mic button next to Send, speak, then release. LakshX transcribes what you said and inserts it at the caret in the composer — it never sends automatically, so you always get to review or edit the text first.
Push-to-talk hotkey
Prefer not to hold your mouse down? Open the AI Providers panel and set a Push-to-talk hotkey under the voice section. Unlike the mic button, the hotkey is a toggle: press it once to start recording, press it again to stop — a bare key-release is too unreliable to catch consistently, so a toggle sidesteps that. It only fires while the LakshX panel itself has focus, not from an editor tab.
- Any function key (
F1–F24) works on its own. - Anything else needs a modifier —
Ctrl,Alt, orCmd/Win— so the binding can never collide with normal typing.
Fully local, first-use download
Transcription runs on your machine via a bundled Whisper model (base.en) — no cloud speech API, no account, no per-use cost. The first time you use it, LakshX downloads the model (about 142MB) into ~/.lakshx/models and caches it there; every recording after that is transcribed offline. Recognition is seeded with a prompt biased toward code and developer vocabulary, so identifiers, library names, and commands come through more cleanly than a general-purpose dictation engine.
If the mic button says it can’t get access
Voice can be turned off
lakshx.voice.enabled to false in your VS Code settings.