Mixed-language dictation
Talk through prompts with Chinese, English, and technical terms in one sentence. VoiceBabel turns that into text you can send anywhere.
Free, open-source macOS menu bar tool
Multilingual Voice Input for AI Chats
Speak naturally in mixed languages, then send clean text into ChatGPT, Claude, Gemini, Cursor, VS Code, or any AI input box.
Public release is coming soon.
- Free
- Open source
- Menu bar
The gap
AI workflows became multilingual. Voice input did not.
Developers often mix a native language with English API names, errors, prompts, and code terms. Many dictation tools still expect one language at a time, and some AI products still make English voice input the easy path.
Core workflows
Speak the way you think.
Speak in your language, output English
When an AI tool works better in English, speak naturally and let VoiceBabel produce English text for the input box.
Reliability
Local and cloud engines, with fallback.
A long spoken prompt should not vanish because one engine failed. VoiceBabel can use local WhisperKit and OpenAI cloud transcription, with user-controlled priority and fallback.
Local engine
Offline after the model is ready.
Cloud engine
High-quality OpenAI transcription when configured.
Fallback path
Try the next available engine instead of losing the whole input.
Flow
Press Option, speak, keep moving.
-
1
Press Option
Hold Option for push-to-talk, tap for longer dictation, or double-tap for translation mode.
-
2
Speak naturally
Use your normal mix of languages, code terms, product names, and prompt instructions.
-
3
Text appears
VoiceBabel inserts the result into the active input box and can auto-send immediately or after a delay.
Privacy
Local when you want it. Transparent when you use cloud.
In local mode, audio stays on your Mac. VoiceBabel does not intentionally store recordings after processing, and the current app has no analytics or telemetry.
Read the privacy notesWhy VoiceBabel
Not just another dictation app.
Beyond Apple Dictation
Built for mixed-language technical speech, not only clean single-language sentences.
Beyond meeting transcription
Designed for typing into AI chats and developer tools, not recording meetings or making subtitles.
Free and open source
A small productivity tool should not need a subscription to be useful.
FAQ
Questions before release
Does VoiceBabel work offline?
Local transcription can work offline after the WhisperKit model is available. Cloud transcription and cloud translation require network access.
Do I need an OpenAI API key?
Only for cloud features. VoiceBabel stores your OpenAI API key in macOS Keychain.
What macOS version do I need?
Version requirements may vary by feature while the app is in development. Apple Translation requires macOS 15 or later.
Does it store my recordings?
No. Audio is processed for transcription or translation and is not intentionally kept after processing.
Can it translate my speech into English?
Yes. Translation mode transcribes your speech first, then translates the text into the target language.
What permissions does it need?
VoiceBabel needs microphone access to record and Accessibility permission to insert text into the active app.
Will there be Windows or Linux versions?
Not for now. VoiceBabel is focused on macOS first because it is built as a macOS productivity tool.
Download
GitHub Releases, soon.
VoiceBabel is still in development. The public macOS release will be published through GitHub Releases when it is ready.
Support
Help keep VoiceBabel free.
VoiceBabel is a free and open-source project from a one-person company. It has taken months to build, and the AI tools used during development cost money. If it helps your workflow, you can buy me a coffee.
Buy me a coffeeGitHub Sponsors link pending.