Describe bugs, refactors, tests, and coding tasks by voice, then review the text in Claude Code before you submit it. StarWhisper uses a global Windows hotkey and local Whisper transcription, so your spoken brief can become text in the focused terminal or editor field.
Four steps take you from installation to a reviewed task brief.
Download StarWhisper, install it on Windows, and allow microphone access when asked. Start with a local Whisper model. The free tier includes 2,000 words per week.
Right-click the StarWhisper circle, open Settings, choose the Basic tab, and find the Hotkey section. Set a key combination that is comfortable beside your terminal or editor. The same global hotkey can be used in other apps.
Click the terminal prompt or other text field where you want the task description to appear. If a command is still running, wait for the input prompt. StarWhisper types into the active field, so focus decides where the text lands.
Hold the hotkey, explain the task in complete sentences, then release it. StarWhisper transcribes locally and inserts the text. Check file names, symbols, and technical terms before you submit the request. Paste literal code separately when needed.
Use your voice for the explanation, then keep exact code and commands visible.
Describe what you expected, what happened, the steps to reproduce it, and the part of the app you already checked. A complete brief is easier to review than scattered notes.
Dictate the goal, boundaries, existing pattern to follow, and files that should stay unchanged. This gives Claude Code the shape of the work before it proposes edits.
Speak the normal case, edge cases, and acceptance checks. Then paste the exact API names or code snippets that need to remain character-perfect.
Use the focused field to dictate a commit message, handoff note, or change summary. Review names and commands before copying the text elsewhere.
Claude Code work often starts with context rather than code. You need to explain the goal, constraints, failed attempts, and definition of done. Speaking that context can be more natural than stopping to type every sentence. StarWhisper turns the explanation into text in the field you already use, so you can edit it before sending.
The workflow is deliberately simple: focus the input, hold a hotkey, speak, release, and review. There is no special command syntax to learn for dictation. Use ordinary language for the brief, then keep exact snippets, paths, and punctuation-heavy code in a pasted block.
StarWhisper uses the OpenAI Whisper model family through local whisper.cpp processing. In Local Mode, the audio is transcribed on your computer instead of being sent to a transcription server. That is useful when a task brief contains private project context or when you are working without a network connection.
Pro users may choose an optional cloud Whisper fallback, but it is not required for this workflow. Read the offline speech-to-text guide if keeping audio on the device is the main reason you are comparing dictation tools.
StarWhisper supports the Whisper model range from tiny through large-v3. Smaller models can suit a quick task note, while larger models are available when you prefer a different speed and quality balance. Choose the language in the app or use its language options for multilingual work. StarWhisper supports 100+ languages.
Hardware support includes CUDA, Vulkan, and Metal acceleration paths, depending on the computer and operating system. CPU-only operation is also available. Let the model and hardware fit the length of your dictation instead of expecting a short command and a long design brief to need the same setup.
One global hotkey, with a focused text field in each app.
Dictate coding briefs, fixes, and test plans in your editor.
See the product overview, supported platforms, and core features.
Review the paid plan for unlimited dictation and larger model access.
Learn how local Whisper processing works on Windows.