Whisper Dictation for Windows and Mac: Put Text at Your Cursor

Whisper can recognize the words you speak. To use it for desktop dictation, pair it with an app that captures your microphone, runs a compatible Whisper model, and inserts the transcript into the field you're editing. Start with a local editor so you can see whether each step works. OpenAI Whisper models · Voconly models and desktop support

Pick the right workflow for your input

Check that your chosen build supports your operating system and can run the specific model on your hardware. Whisper's smaller English-only variants are an option for English-only speech. Use a multilingual variant for other languages. If you mix languages mid-sentence, judge the result on a line you actually say.

Get one sentence into an editor

1. In the dictation app, choose an available local Whisper model and let it finish downloading. Confirm the app's microphone and text insertion permissions.
2. Open Notepad or TextEdit. Set the cursor on a blank line, select plain transcription, then use your configured shortcut to say, “Ticket four zero seven is ready for review.”
3. Stop the recording. Check that the words land on the correct line and that the ticket number reads 407. If the app shows a transcript but the editor stays blank, check the app's insertion permissions and where the cursor is.
4. Repeat in the program where you normally write.

Compare two models on the job you need

Try the same short message, a name, and a technical term with each model. Track the time from finishing the phrase to seeing usable text and the corrections you make. Use the model that gets your recurring details right with a wait you can tolerate. For local use without an internet connection, download the model first and repeat the test with any online cleanup disabled.