# Whisperstream Whisperstream is private local dictation software for Windows. It captures your voice with a push-to-talk hotkey, transcribes it entirely on your PC using a local AI model, and types the result into whatever app you have focused. Nothing leaves your machine. One-time purchase at $29 USD, no subscription. ## Key facts - Platform: Windows 10, Windows 11 (64-bit) - Pricing: $29 USD, one-time purchase, no subscription - Free trial: 10 minutes of transcription, no account or credit card required - Publisher: Lanreal Technologies Inc., Toronto, Canada - Product page: https://whisperstream.io - Support: support@whisperstream.io - Publisher contact: contact@lanreal.io - GitHub org: https://github.com/lanreal-io - Parent company: https://www.lanreal.io/ ## Canonical pages - https://whisperstream.io/ - https://whisperstream.io/about - https://whisperstream.io/eula - https://whisperstream.io/privacy - https://whisperstream.io/privacy-website - https://whisperstream.io/subprocessors - https://whisperstream.io/terms --- ## What Whisperstream is Whisperstream is built by Lanreal Technologies Inc., an Ontario corporation based in Toronto, Canada. It is a Windows dictation app that lets you speak anywhere you type and see text land in any app. Transcription runs 100% on your PC using a local NVIDIA Parakeet model. There is no cloud transcription API, no audio upload, and no account tied to your voice. The app is a one-time $29 USD purchase. There is no subscription, no recurring charge, and no paid tiers. All features are included from day one. A free trial gives you 10 minutes of transcription with no account or credit card required. ## How it works Whisperstream uses a push-to-talk hotkey (default: Right Shift). Hold the key while speaking, release to transcribe. The transcribed text is typed directly into whichever window you have focused, so it works in any app: browsers, editors, terminals, chat clients, email, and more. Under the hood, speech is processed by NVIDIA's Parakeet TDT v3 model running via ONNX Runtime on CPU. No GPU is required. The model runs locally in your computer's memory. Here is exactly what happens on each dictation: 1. Your microphone captures audio. 2. The local NVIDIA model running on your machine converts it to text. 3. The text is typed into your active window. 4. The audio is discarded from RAM. Nothing is saved. Nothing is sent. The only network calls Whisperstream ever makes are for license verification and app updates. None of these carry your voice, your transcribed text, or any record of what you dictated. ## Privacy posture Whisperstream cannot access your audio after transcription completes because the audio is never stored and never transmitted. Transcription happens entirely in local memory. - No audio is uploaded to any server. - No transcripts are stored by Whisperstream or sent to the cloud. - No analytics collect transcript content. - No account is required to use the app after purchase. - The local model file is downloaded once during setup and then runs offline. Local processing is not a policy statement. It is how the transcription works. There is no cloud transcription API in the product. ## Pricing Whisperstream costs $29 USD as a one-time purchase. There is no subscription, no annual renewal, and no upsells or paid tiers. All features are included. A 30-day money-back guarantee applies, no questions asked. Contact support@whisperstream.io to request a refund. A free trial is available: download Whisperstream and get 10 minutes of transcription free, with no account and no credit card. For comparison, cloud-based dictation apps typically cost $150 or more per year. Whisperstream's total cost over three years is $29. ## Publisher Whisperstream is published by Lanreal Technologies Inc. - Legal name: Lanreal Technologies Inc. - Address: 18 King Street East, Suite 1400, Toronto, ON M5C 1C4, Canada - General contact: contact@lanreal.io - Product support: support@whisperstream.io - Website: https://www.lanreal.io/ - GitHub: https://github.com/lanreal-io ## Common questions ### Does Whisperstream send my voice to the cloud? No. Whisperstream processes speech locally using NVIDIA's Parakeet model running entirely in your computer's memory. Your microphone audio is converted to text on your PC and then discarded from RAM. No audio is uploaded. No transcripts leave your machine. The only network calls the app makes are for license verification and app updates, and neither carries voice data or transcript content. ### What does it cost? $29 USD, one-time purchase. No subscription. No annual renewal. All features included. A free trial gives you 10 minutes of transcription with no account or credit card required. There is also a 30-day money-back guarantee. ### Does it work offline? Mostly yes. You need an internet connection once to download the speech model (roughly 600 MB) and once to activate your license. After that, Whisperstream runs fully offline. No internet connection is needed for day-to-day transcription. ### What languages are supported? Whisperstream supports 25 languages: English, Spanish, French, German, Italian, Portuguese, Russian, and Polish, plus 17 more: Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, Greek, Hungarian, Latvian, Lithuanian, Maltese, Romanian, Slovak, Slovenian, Swedish, and Ukrainian. ### Is there a Mac or Linux version? No. Whisperstream is Windows-only (Windows 10 and Windows 11, 64-bit). There is no Mac or Linux version at this time. ### What model does Whisperstream use? Whisperstream uses NVIDIA's Parakeet TDT v3 speech recognition model, running via ONNX Runtime on CPU. No GPU is required. The model runs locally on your machine. ### What are the system requirements? Windows 10 or Windows 11, 64-bit. 8 GB RAM minimum, 16 GB recommended. Works on CPU, no GPU required. A modern processor (Intel 8th gen or later, AMD Ryzen 3000 or later) gives near-instant transcription.