Speak, and it appearswhere your cursor is.
Hold the shortcut, talk, let go. Speech recognition and cleanup run on your PC, and the result is typed straight into the app you are already using.
Right Althold to talk
Free to start · No account needed · Works offline once models are downloaded
Holding Right Alt and speaking brings up the recording capsule at the bottom of the screen, which shows your words as you say them. When you let go, the sentence is cleaned up and typed at the cursor in the notes window. The example repeats.
From quick notes to meeting minutes, all on your PC.
Everything below is free and, with the default settings, runs on your own machine.
Type by voice anywhere
Works wherever there is a cursor: notes, browsers, code editors, chat apps. Press the shortcut twice to keep talking hands-free.
Text cleanup
Removes filler words and tidies up sentences with a local Ollama model. You can also create your own commands, such as translate or summarize.
Next-sentence suggestions
While you write, a local model offers ways to continue. Pick one with Ctrl+Alt+Up/Down and insert it with Ctrl+Alt+Enter.
Live captions
Shows captions on top of your screen for meetings, lectures, and videos. Finished lines are corrected once more using the lines around them.
Meeting notes
Records long meetings and, when they end, writes up a summary and action items.
File transcription
Drop in an audio or video file and get a transcript with timestamps.
Hold, speak, release. Done.
No switching windows. It all happens where you were already typing.
Hold
Recording runs while you hold Right Alt. You can change the shortcut in Settings.
Recognize
The built-in faster-whisper engine turns speech into text. An NVIDIA GPU makes it faster.
Clean up
If enabled, a local model tidies or translates the text. Turn it off to keep exactly what was heard.
Type
The text is pasted at your cursor. Whatever was on your clipboard is put back afterwards.
By default, it stays on your PC.
There are cloud features, but they are not used until you turn them on.
- Speech is processed on your PC
- The default speech recognition uses a local engine, and recordings and history are stored in a database on your PC.
- No account required
- Every local feature works without signing in.
- Usage analysis is off
- Typing-pattern analysis only turns on with your consent, and even then the results stay on your PC.
- The cloud is opt-in
- If you turn on cloud speech recognition (with your own API key) or cloud text cleanup (sign-in required), only those requests are sent to an outside server.
Every local feature is free.
Paid plans let you use cloud text cleanup more often and with larger models.
Free
Everything that runs on your PC
₩0
- Unlimited dictation, captions, meeting notes, and file transcription
- Unlimited cleanup with local models
- Cloud cleanup 250 times a week (sign-in required)
Pro
If you use cloud cleanup every day
₩2,900/mo
- Everything in Free
- Cloud cleanup per day: standard model 1,500 times
- Advanced model 300 times · top model 50 times
Pro+
Plenty of room for large models
₩8,900/mo
- Everything in Pro
- Cloud cleanup per day: standard model unlimited
- Advanced model 1,500 times · top model 300 times
Prices are monthly, in Korean won. Payment and cancellation are managed on your web account page.
Download
Once installed, new versions are installed automatically.
Windows v1.9.0
Windows 10 · 11 (64-bit)
Published September 27, 2026 · The speech model is downloaded the first time you run the app.
Release notes(opens in a new tab)File hashes (latest.yml)(opens in a new tab)
Coming later
macOS (Apple Silicon)
Will be posted here once signing and installation checks are complete.
Android · iOS
Will be posted here once store review is complete.
D3RO Voice at a glance
- What it is
- A voice typing app for Windows. Hold a shortcut, speak, and the text is typed at the cursor of the app you are using.
- Platform
- Windows 10 · 11 (64-bit)
- Price
- Every local feature is free. For heavier cloud text cleanup, Pro is ₩2,900/mo and Pro+ is ₩8,900/mo.
- Internet
- Once the speech model and a local AI model are downloaded, dictation and cleanup work offline.
- Speech recognition
- faster-whisper runs on your PC. It detects the language automatically, or you can pick Korean, English, Japanese, or Chinese.
- Text cleanup
- A local Ollama model removes filler words and tidies sentences. Cloud models are optional.
- Your data
- Recordings and transcripts are stored in a database on your PC. No account is required.
- Latest version
- 1.9.0 (published September 27, 2026)
Frequently asked questions
Not for dictation: the speech engine is built into the app. To run text cleanup and translation on your PC you need Ollama, and the app walks you through installing it.
No, it runs on the CPU. An NVIDIA GPU (CUDA) makes recognition faster.
Yes. Once the speech model and an Ollama model are downloaded, dictation and cleanup work offline. Paid plans occasionally go online to confirm the subscription, and keep working for up to 30 days without a connection.
Most Windows apps where you can type. Text is inserted by pasting, so fields that block pasting may not accept it.
Every feature that runs on your PC is free with no usage limits. Paid plans raise how much cloud text cleanup you can use.
Only Windows is available right now. The macOS (Apple Silicon) and mobile apps will be posted here once they pass their checks.
By default the spoken language is detected automatically. You can also choose Korean, English, Japanese, or Chinese in Settings.
In a database on your PC; they are not sent anywhere. Only if you turn on cloud speech recognition or cloud text cleanup are those requests sent to an outside server.