For developers · Windows · free

Voice dictation that knows it's code.

Most dictation tools punctuate getUserById into "Get user by ID.". VoxWave knows when you're in VS Code, Cursor, Claude Code or a terminal, and switches to a code-aware cleanup profile. Free, unlimited, runs offline.

Download for freeSee it on GitHub →
01 — the code-aware profile

Detects your editor. Adapts the cleanup.

The window-detector matches the active executable against a known list — VS Code, Cursor, Claude Code, Windsurf, JetBrains, Sublime, Notepad++, Zed, Fleet, Lapce, Emacs, Vim/Neovim, plus every common terminal (Windows Terminal, cmd, PowerShell, bash, zsh, fish, Konsole, Alacritty, WezTerm, Kitty, Rio, Ghostty).

In any of those, cleanup switches to code mode: no period after every sentence fragment, no auto-capitalization of variable names, no fancy quotes that break your strings, no em-dashes inside an arrow function.

Ctrl+Shift+V
default hotkey · remap to whatever you want
report.ts — voxwave-app
1
2
3
4
5
function sendReport(date) {
// dictated: 'send the daily report for today'
const subject = `Daily report — ${date}`;
return mailer.send({ subject });
}
voxwave · code-aware profile
02 — works everywhere you code

Inject anywhere. No editor plugin required.

VoxWave doesn't talk to APIs or extensions — it injects keystrokes the same way your keyboard does. That means it works in every editor and every terminal that accepts text input. No plugin to install, no language server to configure, no permissions to grant.

Editors
VS Code · Cursor · Claude Code · Windsurf · JetBrains · Sublime · Notepad++ · Zed · Fleet · Vim / Neovim · Emacs
Terminals
Windows Terminal · cmd · PowerShell · bash · zsh · fish · Konsole · Alacritty · WezTerm · Kitty · Ghostty
AI IDEs
Cursor · Claude Code · Windsurf · GitHub Copilot Chat · Continue.dev — all detected as code profile
Notes & docs
Notion · Obsidian · Logseq · Word · LibreOffice · Google Docs — switches to prose cleanup automatically
03 — privacy if you want it

Local mode keeps your code on your machine.

Don't want your prototype code passing through Groq or OpenAI? Switch to local mode: Settings → Advanced → Transcription provider → local. Whisper runs on your CPU or GPU via faster-whisper, and the AI cleanup runs through a local Ollama model (or skips entirely with regex). Zero outbound traffic, zero API key needed.

Cloud mode is the default because it's faster (≤3s for a 10s dictation) and the quality is better. You decide.

04 — questions devs ask

Honest answers, no marketing.

How does VoxWave avoid breaking camelCase or snake_case?

The AI cleanup is context-aware — it detects the active editor (VS Code, Cursor, Claude Code, JetBrains, Vim, terminals) and switches to a 'code' profile. In that profile it skips punctuation rewrites, preserves identifier casing, and never adds periods inside function names or variables. You dictate "camel case send report" and you get sendReport, not Camel case send report.

Can I dictate code comments in Python or JavaScript?

Yes — that's actually where it shines. Hit your hotkey at the start of a line, dictate '# this function handles the daily report', and you get a clean comment with proper grammar. Same for // in JS/TS/Rust/Go.

Does it work inside Claude Code, Cursor and Windsurf?

Yes. VoxWave injects text via the clipboard (Ctrl+V), which works in every Electron editor including Claude Code, Cursor, Windsurf and the regular VS Code. It also detects the executable name so the cleanup profile adapts automatically.

Does it work in terminals (WT, cmd, PowerShell, WSL)?

Yes. Terminal detection is built in — Windows Terminal, cmd, PowerShell, bash, zsh, fish, Konsole, Alacritty, WezTerm, Kitty, Ghostty are all recognized. In a terminal, cleanup goes into 'code' mode by default (no punctuation rewrites) so you can dictate commands literally.

Can I use it offline so my code never hits a cloud?

Yes. Switch to local mode in Settings → Advanced → Transcription provider: local. Whisper runs on your CPU or GPU via faster-whisper, the cleanup runs through a local Ollama model or pure regex. Zero outbound traffic, no key required.

What's the latency for a 10-second dictation?

On hybrid mode (default) you get full cleaned output in under 3 seconds end-to-end on a 50 Mbps connection. Local mode is slower (5-10s) but private. Progressive injection drops the raw transcript instantly (~200ms) and silently replaces it with the cleaned version when ready, so you can keep typing.

Is there a Vim or Neovim integration?

VoxWave injects keystrokes, not API calls — it works in Vim/Neovim just like typing would. Press the hotkey in normal mode to enter insert mode and dictate, then exit. There's no editor plugin to install.

Talk to your editor.

Free forever. No account. ~28 MB. Works offline.

Download VoxWave