Summary
A free Mac dictation app that transcribes speech locally and inserts text into applications.
About
WhisperWrite applies local speech recognition to cross-application text entry on Mac, combining offline operation with optional AI-assisted editing.
WhisperWrite is a dictation app for Mac. Users hold a key and speak to enter text in applications such as Mail, Slack, Google Docs, ChatGPT, and VS Code. Speech recognition runs locally using an NVIDIA Parakeet-based model, works offline, and is described as producing text in about 0.1 seconds. The app removes filler words, adds punctuation, and learns names and jargon. It has no word limits, subscription, or account requirement. Optional AI editing is available through Claude using the user's own API key.
Key Features
Hold-to-dictate text into supported Mac applications
On-device speech recognition using an NVIDIA Parakeet-based model
Offline operation with voice data remaining on the computer
Automatic removal of filler words and addition of punctuation
Personalization for names and jargon, with optional Claude editing using the user's own key
Real-World Use Cases
Drafting messages and documents
Dictate emails, chat messages, and documents directly into applications such as Mail, Slack, and Google Docs.
Developer notes and prompts
Enter spoken notes, instructions, or prompts into tools such as VS Code and ChatGPT.
Offline text entry
Use speech-to-text on a Mac without an internet connection, with voice processing performed locally.
Business Opportunities
Reduce routine writing time
Problem: Professionals spend time typing routine messages, notes, and drafts across multiple applications.
Solution: Use cross-application dictation with automatic punctuation and filler-word removal to accelerate initial text entry.
Estimated ROI: Potentially positive for frequent writers if dictation reduces time spent on routine text; actual returns depend on usage and accuracy.
Execution: Introduce the app as an optional Mac workflow tool, identify recurring dictation tasks, and assess time saved and transcription quality in those workflows.
Support privacy-conscious voice workflows
Problem: Some users avoid cloud-based transcription because they do not want voice data sent off-device.
Solution: Use local speech recognition and offline operation for dictation, while keeping optional Claude editing separate and dependent on the user's own key.
Estimated ROI: Potentially valuable where local processing supports adoption, though financial impact depends on organizational privacy requirements and user uptake.
Execution: Evaluate local dictation with non-sensitive workflows first, and establish organizational guidance for any use of optional external AI editing.