Push-to-talk
A global hotkey you choose. Hold to record, release to stop — text is typed straight into any app: editors, chat, email, IDE.
Open source · AGPL-3.0 · under 15 MB
Speak. It's written.
Hold a key, speak, release — text lands at your cursor. Your API key, your cloud, zero subscriptions.
Push-to-talk · global hotkey
Bring your own key · pay at cost
macOS · universal DMG
Windows · MSI installer
under 15 MB native binary
EN / RU interface
Push-to-talk · global hotkey
Bring your own key · pay at cost
macOS · universal DMG
Windows · MSI installer
under 15 MB native binary
EN / RU interface
Everything dictation needs. Nothing it doesn't.
A tiny native app that stays out of your way.
Bring your own key
Paste your own API key and pay providers at cost. No middleman, no subscription, no usage caps invented by us.
sk-…e9f2 → stored in Keychain / DPAPI
13.7 MB
Featherweight native
Built with Tauri + Rust: an under 15 MB binary, instant start, negligible RAM. No Electron, no tray-resident bloat.
Your audio, your rules
Audio goes only to the provider you configured. No telemetry, no accounts, no analytics — the code is open, check it.
macOS & Windows
One mental model on both desktops, localized interface in English and Russian, DMG and installer for your platform.
macOS 12+ · DMG
Windows 10/11 · MSI
Your voice, your cloud, zero subscriptions.
Three seconds from speech to text
No windows to open, no files to manage.
Hold the hotkey
Press and hold your chosen shortcut in any application. The menu-bar indicator lights up.
Speak naturally
Dictate a sentence, a paragraph or a commit message. Punctuation by voice, too.
Release — done
Transcribed text is inserted at your cursor. That's the whole workflow.
Bring a key, pick a brain
Configure one or all three. VoiceDo never proxies your traffic.
Default
OpenAI-compatible
Whisper and any endpoint that speaks the OpenAI audio API — including self-hosted servers on your LAN.
POST /v1/audio/transcriptions
Best for RU
Qwen · DashScope
Alibaba Cloud Qwen speech models — strong Russian and multilingual recognition at low cost.
qwen3-asr · paraformer
No key needed
Google Speech
Zero-config fallback using Google's public transcription endpoint. Great for trying VoiceDo in one minute.
speech-api · instant
Type less. Say more.
Free, open source, under 15 MB. Grab the build for your platform and dictate your first sentence in under a minute.
Demo builds are not digitally signed yet — macOS: right-click → "Open" on first launch; Windows SmartScreen may warn "unknown publisher", use "More info → Run".
Like it? Keep it free.
VoiceDo is open source and always will be. Donations fuel development.