Whisperstream

Whisperstream is a local AI dictation app for Windows. Pay once, voice stays on your machine. Press a push-to-talk hotkey, speak, and your transcribed words paste into whatever window has focus. Transcription runs on your CPU via NVIDIA Parakeet TDT. No cloud upload, no third-party API, no model trained on your voice. There's nothing to send because there's nowhere to send it. $29 one-time, lifetime updates. No subscription, no per-minute fees, no usage caps. Subscription cloud dictation services charge $100-150 per year and process your speech on their servers; Whisperstream charges $29 forever and processes it on yours. Key features: - 100% local, on-CPU transcription. No network calls during dictation. - $29 one-time purchase. Lifetime updates. 30-day money-back guarantee. - Free trial on first install, no account required. - Push-to-talk hotkey or toggle mode. Bind to any key. - Custom word replacements for jargon, names, code, acronyms. - Auto-lowers other audio while you dictate. Built for: - Privacy-critical professionals (legal, healthcare, defense, NDA-bound engineering) where cloud dictation fails compliance and training-data clauses are deal-breakers. - Windows builders dictating into Cursor, Claude Code, VS Code, and JetBrains IDEs who don't want code-related speech going to a third-party API. - Accessibility and RSI users who need a hotkey-and-speak workflow without learning a scripting language. - Writers and content creators drafting by voice who don't want their drafts indexed. - Academics and ESL users who think faster than they type and want their speech kept private. Tech stack: NVIDIA Parakeet TDT v3, ONNX Runtime, Tauri v2, WebView2. Made by Lanreal Technologies Inc.