Whisperstream is a local voice dictation software for Windows, with the core goal of letting users replace keyboard input with speech. It is officially positioned as "private dictation," emphasizing that transcription is completed locally on the device and microphone audio is not uploaded.
Uses and scenarios It is suitable for users who want to reduce typing and increase text entry speed. Officially, typing speed is typically 40 to 60 words per minute, while thinking and speaking speed is about three times that. Dictated text is output at the current cursor position, so it can be used in existing applications such as Outlook, VS Code, and browsers without installing add-ins. The official materials also list offline scenarios, such as use on airplanes, trains, or when Wi-Fi is interrupted.
Main capabilities
- Local transcription: The core speech recognition and text output can run offline after setup, and audio is not uploaded; officially, the core transcription path has no cloud transcription API or audio upload.
- Global hotkeys: Any key can be configured, supporting push-to-talk or toggle hands-free mode.
- Application compatibility: Use it in existing applications without add-ins.
- Local text cleanup: Local models can remove filler words and verbal tics, and apply the user's spoken corrections; online enhancement providers can also be configured.
- Custom dictionary: Add names, terms, and abbreviations once.
- Per-application switching: Custom AI prompts can be written for each application or website and applied automatically.
- Transcription history: Past dictations can be searched and audio replayed, with records stored encrypted on the device.
- Import audio files: Recording files can be imported and transcribed, all locally on the machine.
- Language support: Supports 39 languages. Officially listed are English, Spanish, French, German, Italian, Portuguese, Russian, Polish, Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, Greek, Hungarian, Latvian, Lithuanian, Maltese, Romanian, Slovak, Slovenian, Swedish, Ukrainian, Arabic, Cantonese, Chinese, Filipino, Hindi, Indonesian, Japanese, Korean, Macedonian, Malay, Persian, Thai, Turkish, and Vietnamese. All run locally, with no cloud required.
How to use Official describes the dictation process: the microphone captures audio, a local NVIDIA model converts it to text, the text is entered into the currently active window, the audio does not reach its servers, and saved recordings are stored encrypted on the computer. Core transcription does not make network calls or upload audio. Separate network connections are used for license checks, updates, model downloads, voluntary feedback, Agent Speak's Claude Code workflows, as well as user-configured online enhancement providers. Only after the user enables and configures a cloud enhancement path will transcribed text be sent to the cloud enhancement provider.
Pricing and Licensing One-time payment, no subscription. $29 activates 1 device; $49 activates up to 2 devices you own, such as a laptop and a desktop. Includes all features, no account required, and comes with a 30-day money-back guarantee. You can try it for free, with no account or credit card required; at least 7 days from the first dictation that produces text, and eligible versions may include a temporary extension, with the specific end date viewable in Open Pro in the updated app.
System Requirements and Limitations Requires Windows 10 or 11 (64-bit), minimum 8GB of memory, 16GB recommended, and can run on CPU without a GPU; newer processors (Intel 8th generation and above / AMD Ryzen 3000 and above) can deliver a near-instant transcription experience. Core speech recognition and text output can run offline after setup, but license verification, updates, some model downloads, optionally configured cloud cleanup, and external destinations such as Featurebase or the model services used by Claude Code still require an internet connection.
Support and Comparisons Product questions or bug reports can be sent by email to support@whisperstream.io. For a 30-day refund, please read the full refund terms and send an email from the address used at checkout. Official comparisons are also available with Wispr Flow, Aqua Voice, OpenWhispr, and Win+H, as well as a summary of alternatives such as Dragon, MacWhisper, and superwhisper, along with guide pages for Windows dictation software, offline dictation software recommendations, Claude Code voice input, and more.
