01 / SET UP
Choose your shortcut.
Download a local speech recognition model, check your microphone and choose a comfortable key for dictation.
Matraca 2.0Windows · macOS availableFrom thought
From thought
to text.
Your ideas move at the speed of your voice.
Dictate messages, notes and prompts without stopping to type. Matraca turns your speech into text wherever you need it.
Windows x64 · Open source · macOS available
Downloads and updates on GitHub. See releases and installation instructions.
The Windows installer is not yet signed and may trigger a system warning.
01 / YOUR VOICEAn idea in motion
One sentence. No typing.
Watch your voice turn into text.
02 / YOUR TEXT
Shall we meet tomorrow at nine?
Illustrative example. No microphone access.
Speech recognition on your computerOne shortcut to get startedNo account. No telemetry.
Fits the way you workKeep your
Keep your
train of thought.
Matraca stays in your system tray, ready when you need it. Keep using your editor, browser or chat. Your voice becomes part of your existing workflow.
02 / SPEAK
Let the idea out.
Place your cursor where you want the text and speak. Record a whole message or deliver sentences as you pause.
03 / KEEP GOING
Your text follows along.
Review the text and carry on. Need to recover a sentence? Find it in your local history, when enabled.
Not just for developersYour day has
Your day has
plenty to say.
Matraca started as a way to dictate prompts. The same idea now helps with the everyday writing that fills your day.
That longer message.
Explain the context at your own pace, without turning every thought into a typing marathon.
An idea worth keeping.
Capture notes, organize your thoughts and get a first draft down before you lose the thread.
A clearer request for AI.
Dictate instructions and context into your terminal or work tools. No need to cut your explanation short just to type less.
Your voiceCaptured in memory
Your textLocal Whisper
Audio is never uploaded or saved to disk. History and technical logs may contain transcribed text, stored locally.
Local by designYou speak.
You speak.
Your audio stays.
Speech recognition runs on your computer. Once the model is ready, you can dictate offline.
AI text review is a separate choice. It is off by default. If you enable an external provider, the transcribed text is sent to that service, subject to its terms and usage charges.
Learn about privacy and connectionsBefore you startNo fine
No fine
print.
What to know about this release.
Is Matraca free?
Matraca is available under the MIT license, and the app is free to download. Local speech recognition does not require a paid API. If you enable text review through an external AI service, that provider may charge for usage.
Does it work on Windows and Mac?
A public installer is available for Windows x64. A public Apple Silicon macOS package is available on the GitHub releases page. It is signed only with the local development identity "Matraca Dev"; it has no Apple distribution certificate or notarization. Automatic updates are not available yet. No Linux release has been announced.
Do I need an internet connection to dictate?
You need internet access to download the app and the speech recognition model. After that, local transcription can work offline. Remote AI review and balance checks require a connection when used.
Why might Windows show a warning?
This version of the installer is not yet code-signed. Check the file source, release notes and published SHA-256 on the official release page before deciding to install it. This page does not ask you to disable Windows protection.
Can it type into any app?
It is designed to insert text into editors, browsers, chats and terminals. Compatibility depends on the field and window permissions. If insertion fails, you can copy the message from history, when enabled.
How are AI review API keys stored?
In the current version, keys entered in the interface are stored in the local configuration file without encryption. You can also configure them through environment variables. No API key is needed for local dictation alone. See the review documentation.
Your next idea can start out loud.
Explore the project, follow new releases and contribute.