Local speech transcription
Converts speech to text on the Windows PC using NVIDIA's open-weight Parakeet model on the CPU. After the model download, core dictation can work without an internet connection.
Whisperstream
Freemium
Whisperstream is a Windows dictation app that transcribes speech locally and inserts the resulting text into compatible apps. It is designed for people who want push-to-talk voice input without sending microphone audio to a cloud transcription service.
O que é Whisperstream?
Whisperstream is a Windows dictation toolkit for x64 Windows 10 and Windows 11 computers. It captures speech through a configurable push-to-talk shortcut, transcribes it on the computer with NVIDIA's Parakeet model running on the CPU, and delivers the text to the focused field through clipboard paste or optional simulated keystrokes. The core transcription path does not upload, stream, or store microphone audio on a server, and it can operate offline after the required speech model has been downloaded.
The app is intended for entering text in many existing workflows rather than requiring a dedicated editor or add-in. The site lists Outlook, Word, Slack, VS Code, browsers, terminals, Gmail, and Claude Code among compatible or demonstrated destinations, while noting that application, privilege, security, administrator, and remote-session policies can affect text insertion.
Beyond basic dictation, Whisperstream provides transcript history, audio-file transcription, a custom dictionary, optional AI cleanup, and application- or website-based profiles. Cleanup can use a local model on supported hardware, Ollama, or a configured cloud or custom provider; it is off by default. The app supports 39 languages, can lower system volume while recording, and includes Lockdown Mode to restrict cleanup to an on-device model and turn off Whisperstream cloud enhancement.
Whisperstream is available as a free trial without an account or credit card. The listed purchase options include a $29 one-device license and a $49 license for up to two of the buyer's own devices, with a 30-day refund policy stated on the site.
O que Whisperstream pode fazer?
Push-to-talk controls
Hold the default Right Shift shortcut to dictate, remap the shortcut, or switch to toggle mode for hands-free recording. The app does not require administrator rights to register the shortcut.
Text delivery across applications
Inserts dictated text into a focused field by clipboard paste or optional simulated keystrokes, without requiring an add-in. Supported behavior can vary by application and security or remote-session policy.
Multilingual models
Supports 39 languages, with the app selecting and downloading the matching on-device speech model when needed.
History, file import, and dictionary
Search saved transcripts, replay associated audio when available, copy results again, transcribe selected audio files locally, and add spelling overrides for names, technical terms, and acronyms.
Optional cleanup and privacy controls
Remove filler words, correct punctuation, and apply spoken corrections using a local or configured provider. Lockdown Mode disables Whisperstream cloud enhancement and limits cleanup to an on-device model.
Casos de uso
“Writing email and documents”
Dictate into Outlook, Word, Gmail, or another focused text field, then use optional cleanup for punctuation, grammar, and spoken corrections before sending or saving.
“Coding and terminal work”
Enter text in VS Code or terminals, with a custom profile able to apply a terse or otherwise task-specific cleanup prompt for the focused application.
“Chat and collaboration”
Use push-to-talk dictation in Slack and browser-based tools without installing an application add-in, subject to the destination's text-input policies.
“Offline or privacy-sensitive work”
Dictate while traveling or when connectivity is unavailable after setup, or use Lockdown Mode when cloud enhancement should remain disabled and local history requires additional protection.
“Transcribing existing recordings”
Import an audio file stored on the computer and generate a transcript locally, with optional saving to searchable history.
Perguntas frequentes
What operating systems and hardware does Whisperstream require?
It supports Windows 10 and Windows 11 on x64 computers. The stated minimum is 8 GB of RAM, with 16 GB recommended; core transcription runs on the CPU and does not require a GPU. Intel 8th-generation or newer and AMD Ryzen 3000-series or newer processors are described as providing near-instant transcription.
Does Whisperstream upload microphone audio?
The core transcription path keeps microphone audio on the PC and does not upload it to Whisperstream servers. Optional online features, configured cloud enhancement, license checks, updates, model downloads, and separate integrations have their own network activity; configured cloud enhancement receives transcript text rather than the original voice recording.
How does text reach another application?
Whisperstream sends the result to the focused field using clipboard paste or optional simulated keystrokes. It can work in applications such as Outlook, Word, Slack, VS Code, browsers, and terminals, although application permissions, security settings, administrator restrictions, and remote-session policies may limit insertion.
Is there a free trial and what happens after purchase?
The app can be tried without an account or credit card and provides at least seven days from the first dictation that produces text. The site lists $29 for one device and $49 for up to two devices owned by the purchaser, plus a 30-day refund policy.
Can I use Whisperstream without an internet connection?
Core recording and transcription can run offline after setup and the required model download. Internet access may still be needed for activation or periodic license validation, updates, model downloads, configured cloud enhancement, and certain separate workflows.
Informações rápidas
- Category
- Windows dictation and productivity software
- Platform
- Windows 10 and Windows 11, x64
- Transcription
- On-device CPU processing with NVIDIA Parakeet and additional language models
- Language support
- 39 languages
- Primary workflow
- Push-to-talk speech converted into text in a focused application field
- Pricing model
- Free trial; $29 one device or $49 for up to two personal devices; 30-day refund policy
Análise de tráfego de Whisperstream
Os dados de tráfego são apenas para referência.
- Classificação do domínio
- 16
Alternativas ao Whisperstream
Voiser
voiser.net
AI tools for converting text to speech, transcribing audio and video, and dubbing YouTube content.
Blurt
www.assemblyai.com
Blurt is AssemblyAI’s voice-to-text dictation app for entering spoken words into the field or text destination where you are working. Hold a key, speak, and release it to send the dictated text onward.
Cartesia
cartesia.ai
Speech models and managed voice agents for teams building real-time voice experiences.
Aqua Voice
withaqua.com
Aqua Voice converte voz em texto em tempo real no Mac, Windows, iPhone e em fluxos de trabalho de IA, com API para equipes e desenvolvedores.
VoiceOS
voiceos.com
VoiceOS é um sistema operacional por voz e assistente de IA para Mac e Windows, com ditado, mensagens, buscas na web e automação de apps.
Speakly
speakly.ai
Speakly é uma ferramenta gratuita de voz para texto com IA que transforma fala em texto refinado em mais de 100 apps, com edição por IA, atalhos e modo agente para Mac, Windows, iPhone e Android.