Instant, accurate voice-to-text transcription at your fingertips. Works offline with local AI or connects to OpenAI for ultimate accuracy.
Free forever - 500 words/day included. No account required.
Powerful features designed for seamless voice transcription
Works completely offline with local Whisper models. No internet required.
Your voice and text never leave your device in offline mode. Complete privacy guaranteed.
Ultra-low latency transcription with GPU acceleration support for NVIDIA cards.
Choose from models between tiny and large based on your accuracy needs.
Floating widget stays on top of all windows. One tap to start dictating.
Choose from 14 preset themes or create your own with custom gradient colors for every state.
See words appear live as you speak. Floating text overlay updates in real-time with adjustable chunk size.
Set any global keyboard shortcut to toggle recording from any app. Default: Ctrl+Space.
Drag & drop audio or video files (MP3, WAV, MP4, MKV) for instant transcription. Pro feature.
Say "new paragraph", "comma", "period", or "backspace" while dictating. Natural voice-driven formatting.
Supports 29+ speech languages with auto-detection. App interface in 11 languages.
Add names, brands, and technical terms. Fuzzy matching auto-corrects variations. Supports exact replacement rules.
Expand short phrases into full text blocks. Say "sig" to insert your complete email signature.
Vocabulary, macros, and settings sync across all your devices. Sign in with Google or email.
Transcribed text is automatically copied and pasted wherever your cursor is. Zero extra steps.
Polish transcriptions with AI-powered formatting. Email, notes, formal letter, or paragraph styles.
All transcriptions saved locally with timestamps. Search, review, and export as text or markdown.
Get started in seconds
Guided Windows installer package. No complex setup required.
Select offline mode for privacy or API mode for best accuracy.
Click the floating button to begin transcribing instantly.
Start free, upgrade when you need more
No account required
"I dictate patient notes directly into my EMR system. The offline mode means zero compliance worries, and the accuracy with the large model is remarkable."
"I switched from Dragon NaturallySpeaking. StarWhisper is faster, works with any app, and the real-time inline transcription is something Dragon never had."
"The GPU acceleration on my RTX card gives me near-instant transcription. I use it for meeting notes, emails, and even coding comments. Saves me hours every week."
Find answers to common questions about StarWhisper
StarWhisper is a Windows desktop dictation app that turns speech into text and types it directly into any active Windows application using OpenAI's Whisper model via whisper.cpp.
Yes. The free tier and the local mode of the Pro tier transcribe entirely on the user's Windows computer with whisper.cpp. No internet connection is required after the initial download. Optional cloud Whisper fallback is available for Pro users in noisy or low-spec environments.
StarWhisper is available for Windows 10/11 and for macOS on Apple Silicon (M1 or newer, macOS 12+), with Metal GPU acceleration. Linux and mobile (iOS, Android) apps are not available yet.
StarWhisper has a permanent free tier with 500 words per day and 3,500 per week, no account or credit card required. Pro is 10 USD per month or 80 USD per year for unlimited transcription, all Whisper models including Medium and Large, file transcription, and optional cloud fallback. A 30-day refund applies to Pro.
Not by default. The free tier and the local Pro mode keep audio on the user's device and do not store or transmit it. Only when the user explicitly enables cloud Whisper fallback (a Pro option) is audio sent to OpenAI's Whisper API for that single transcription; even then it is processed in-memory and not retained by StarWhisper.
StarWhisper is Windows-first, runs Whisper locally for free, and costs 10 USD per month. Wispr Flow is Mac-first, processes audio in the cloud, and uses a per-month SaaS pricing model. Users who run Windows or who care about audio staying on their device generally choose StarWhisper; users who run macOS and prefer a cloud-based SaaS often choose Wispr Flow.
Yes. Whisper recognizes approximately 99 languages and StarWhisper exposes 30 in the speech-language picker, including Spanish, French, German, Portuguese, Russian, Chinese, Japanese, Korean, Italian, Dutch, Polish, Turkish, Arabic, Hindi, and more. All transcription languages work fully offline. The app user interface is available in 11 languages.
Windows 10 (build 1903 or later) or Windows 11, 64-bit. A CPU from 2017 or later runs the Tiny and Base Whisper models in real time. For larger models, an NVIDIA GPU with CUDA support (GTX 10 series or newer) is recommended but not required. CPU-only operation is supported throughout.
Yes for most modern dictation needs. Dragon costs 200 to 500 USD as a one-time purchase and supports about 6 commercial languages. StarWhisper costs 10 USD per month, supports 99 languages via Whisper, runs locally on Windows, includes NVIDIA CUDA acceleration, and has a free tier. Dragon may still be preferable for very specialized commercial vocabularies with mature accuracy tuning.
StarWhisper Pro can transcribe audio and video files (WAV, MP3, MP4, M4A, and similar) by drag-and-drop. It is not a meeting note-taker with speaker labels; for that use case Otter, Fireflies, or Granola are more appropriate. StarWhisper is primarily a real-time dictation tool that types into any active Windows application.
Join professionals using StarWhisper for faster, more accurate transcription
ZIP installer • Downloads 499 MB–2.4 GB during setup
Windows 10/11 (64-bit) • v1.3.256 • GPU acceleration (NVIDIA / Vulkan)
