Dictate a long prompt into Claude, inspect the text, then send it when it is ready. StarWhisper turns speech into text in the focused Windows input with a global hotkey and local Whisper transcription.
Install once, then use the same focused-field workflow for every prompt.
Download StarWhisper, install it on Windows, and allow microphone access. Use a local Whisper model to start. The free tier includes 2,000 words per week.
Right-click the StarWhisper circle, open Settings, choose the Basic tab, and use the Hotkey section. Pick a combination that does not interfere with your normal browser or app shortcuts.
Open Claude and click the message input where you want the prompt to appear. Check that the cursor is in the intended conversation before recording. StarWhisper sends text to the active field, not to a named integration.
Hold the hotkey, speak naturally, and release it when finished. StarWhisper transcribes locally and inserts the result. Review names, instructions, and formatting, then send the prompt yourself.
Voice is a fast way to get context and intent into a prompt you can still edit.
Dictate the role, background, task, constraints, examples, and desired format in one pass. Read the resulting text before you ask Claude to work.
Speak rough ideas, headings, audience notes, and the order you have in mind. Then ask Claude to organize them without losing your original context.
Describe the behavior you want, the error you saw, and the trade-offs that matter. Paste the exact code separately when spelling and punctuation must stay unchanged.
Dictate the audience, tone, length, facts to preserve, and points to remove. The editable transcript gives you a chance to correct proper names before sending.
A useful AI prompt often needs more than a sentence. It may need context from a meeting, the reason for a decision, examples of an acceptable answer, and limits on what should change. Speaking that material can help you capture the full brief while the thought is still fresh.
StarWhisper keeps the final step in your hands. The transcript appears in Claude's text field, where you can edit it, add a pasted reference, or remove anything that does not belong before you send it.
StarWhisper uses OpenAI Whisper through local whisper.cpp processing. In Local Mode, speech is transcribed on your computer. The audio does not need to leave the device for the text to appear in the focused Claude input.
Pro users can choose an optional cloud Whisper fallback. If local processing is the reason you are evaluating a desktop dictation tool, see the offline speech-to-text guide for the local setup.
StarWhisper supports the Whisper model range from tiny through large-v3 and supports 100+ languages. A short question and a long multilingual brief do not have to use the same model choice. Select a language or use the available language options in the app.
CUDA, Vulkan, and Metal acceleration paths are supported according to hardware and operating system. CPU-only processing is also available. Pick the option that fits your computer and the length of the prompt you plan to dictate.
Use the same global hotkey in other focused text fields.
Dictate coding tasks, tests, and bug context in a terminal prompt.
Speak coding briefs and documentation in your editor.
Capture pages, recaps, and project updates by voice.
See the product overview, platform details, and features.
Review the paid plan for unlimited dictation and larger models.
Learn about local Whisper processing on Windows.