StarWhisper works alongside JAWS, NVDA, and Narrator on Windows. Press a hotkey, speak, your screen reader reads the inserted text back. Runs locally, no upload, free plan with no card.
No competition with JAWS, NVDA, or Narrator. StarWhisper writes text, the screen reader reads it.
One global hotkey starts and stops dictation. No mouse target to find, no on-screen widget to click. Right Alt by default, fully reassignable.
StarWhisper simulates keystrokes into whichever field your screen reader has focus on. Email, browser forms, chat, document editors, code editors, all the same.
JAWS, NVDA, and Narrator read the inserted text the way they read any typed text. No new audio pipeline to learn, no special integration mode.
Local Mode keeps audio on the device. Useful for confidential email, legal work, medical notes, and any setting where uploading recordings is not acceptable.
No internet required for Local Mode. Dictate on a flight, in a power outage, in a building with weak signal, on a desk with no Wi-Fi configured.
500 words per day, 3,500 per week, with no credit card and no trial countdown. Pro is $10 per month for unlimited dictation.
Most blind Windows users already have a strong stack. JAWS or NVDA for screen reading, sometimes Narrator as a backup, a Braille display, a known set of keyboard shortcuts for app navigation, and a small list of trusted apps that have proven themselves over years. Voice input belongs in this stack as a complement, not a replacement. The question is whether a new dictation app respects the existing setup or fights it.
StarWhisper is built to sit next to a screen reader, not replace it. There is no overlay, no spoken interface, no chat-style command surface that demands attention. The user presses a global hotkey, speaks, and StarWhisper writes the transcribed text into whichever field has keyboard focus. The screen reader then reads that text the same way it reads any other text inserted into a field, because that is exactly what just happened. No new mental model.
Practical examples. A JAWS user opens a reply in Outlook, presses the StarWhisper hotkey, dictates a two-paragraph response, releases the hotkey, and lets JAWS read the inserted text back with the standard say all command. An NVDA user fills in a long web form by tabbing to each field, dictating the value, tabbing to the next field, dictating that. A Narrator user opens Microsoft Word, dictates a draft, then uses standard Narrator review commands to review and edit line by line. The dictation tool stays out of the way after the text lands.
For a sighted user, dictation accuracy errors are quick to spot and fix. The eye sweeps over the inserted text, catches the wrong word, the cursor moves, the fix happens. For a screen reader user, every dictation error means listening to the inserted text, finding the word that does not match the intended sentence, navigating back to it with arrow keys or character review, deleting it, and re-dictating or typing the correction. The cost of an error is higher because review is slower. The cost of a string of errors compounds.
StarWhisper uses OpenAI Whisper, which is broadly the most accurate open speech model available, and which is unusually good with accents, non-native English, technical vocabulary, and conversational speech. The practical implication for screen reader users is that a typical dictated paragraph needs few corrections, not many. The savings on review time are direct.
Accuracy is also language-aware. Whisper supports 96+ languages, including non-Latin scripts. A Hindi-speaking JAWS user can dictate in Hindi, a Japanese NVDA user can dictate in Japanese, the screen reader reads back whatever language the operating system is configured for. More on language support lives on the multi-language page.
The National Federation of the Blind (NFB) and the American Foundation for the Blind (AFB) both publish ongoing reviews of Windows accessibility tools and dictation apps. Most of their guidance applies to StarWhisper directly: pick a tool that does not fight your screen reader, prefer keyboard-driven interfaces over mouse-driven ones, prefer local processing over cloud where confidentiality matters, and verify that the tool announces or supports announcement of inserted content.
NFB's Access Technology blog and AFB's AccessWorld magazine are useful background reading for any user evaluating a new tool. For blind users new to dictation, the broad pattern they recommend is: install one tool at a time, test it for a week in real work, then decide whether to keep it. StarWhisper's free plan supports exactly that pattern, 500 words a day is enough for a week of email, short notes, and trial workflows before any payment decision.
Many U.S. state vocational rehabilitation programs cover Windows-compatible assistive software for clients pursuing employment. StarWhisper's $10 per month and $80 per year price is small enough that most personal budgets can carry it directly, and the free plan is genuinely usable for users without that budget. The app does not carry an FDA or AT-certified label, it is a general-purpose dictation tool that blind users have found useful in their stack.
The most common workflow. The user navigates to the reply field with their normal screen reader commands, presses the StarWhisper hotkey, dictates the body of the email, releases the hotkey, then uses screen reader review commands to read the text before sending. The save-to-drafts pattern works the same way it always has, the screen reader reads each draft as the user navigates the Drafts folder.
For longer pieces a common pattern is to dictate in chunks, paragraph by paragraph, with a quick review pass after each chunk. The screen reader reads the inserted text, the user catches any obvious errors, corrects them with keyboard editing commands, and moves on. The total time to a clean draft is usually faster than typing for users whose dictation accuracy is high.
Slack, Microsoft Teams, Discord, Telegram desktop, WhatsApp Desktop, and other chat clients all accept keystrokes from StarWhisper into the message compose field. The user focuses the field, presses the hotkey, dictates, releases, and the message is ready to send. Voice notes and audio messages are a different feature offered by those chat apps and are separate from StarWhisper.
Long signup forms, support tickets, contact forms, and admin panels often have many free-text fields. The pattern is the same: tab to each field, dictate the value, tab to the next. StarWhisper does not need a browser extension because it writes via keystrokes, so any browser that accepts keyboard input works, including Chrome, Edge, Firefox, and Brave.
Cloud dictation tools send the audio of your voice to a remote server for processing. For most users this is fine, for some it is not. Sensitive email, medical correspondence, legal drafts, financial records, and personal journals are all categories where many users prefer that nothing leave the device. Blind users are not different in this preference, and they often have less ability to inspect what a tool is sending where, because the inspection workflows themselves are sighted-centric.
StarWhisper Local Mode runs the Whisper model on your own PC. The audio is captured by the microphone, processed by the local model, turned into text, and inserted into the focused field. The audio is not uploaded, not stored on a remote server, not reviewed by humans, not used for training. The architecture is described in detail on the privacy and offline page and in the offline dictation FAQ.
Cloud Mode is available as an opt-in for users who want the highest possible accuracy on a specific session and are comfortable sending that session's audio to the OpenAI Whisper API. It is not the default. Most blind users we hear from keep Local Mode on for daily work and reserve Cloud Mode for occasional high-stakes documents.
Download from the homepage. The installer is a standard signed Windows installer with named buttons that all major screen readers can read. After install, StarWhisper launches and shows a small setup window. The window contains labeled controls for picking a hotkey, picking a default microphone, and picking a default language. All controls are standard Windows form elements, so tab order works as expected, JAWS, NVDA, and Narrator all read the labels, and there are no mouse-only elements.
The default hotkey is Right Alt, which is unused in most workflows. If Right Alt conflicts with an existing JAWS or NVDA shortcut, change it from the dropdown to any single key, modifier combination, or function key. The microphone dropdown lists every Windows audio input device, including built-in laptop mics, USB headsets, and external mics. Most users get noticeably better dictation accuracy from a dedicated USB headset mic, the kind sold for under $50 by many manufacturers.
After setup, open any text field in any Windows app, press the hotkey, speak, release the hotkey. The text appears in the field. The screen reader reads it the same way it reads any other typed text. That is the entire interaction. The Settings window does not need to be opened again unless you want to change the hotkey, the mic, or the language. For users who want to compare with broader use cases, the writers use case and the students use case cover overlapping workflows.
Other ways voice input lowers the friction of writing on Windows
Reduce eye strain by removing the visual cost of typing. Pairs with Magnifier and ZoomText.
Close the gap between thought speed and typing speed. Free, local, in any Windows app.
Skip the spelling tax. Dictate, screen reader or eye reads back, edit lightly.
One hotkey writes to any Windows text field. No per-app plugin, no per-app permission.