Workstation-grade voice typing.
Zero cloud. Zero latency.
Dictate directly into any Windows application at 160+ words per minute with near-zero latency. Powered by Whisper Large v3 Turbo running 100% on your local GPU : no subscriptions, no accounts, and zero audio transmitted over the internet.
Designed for flow. Zero focus theft.
Hardware-accelerated Win32 layered overlays with per-pixel alpha transparency. Receive live visual audio telemetry without stealing cursor focus from your active document, IDE, or browser.
Aura Orb
Radial BloomMinimalist circular badge with radiant vocal intensity bloom and an outlined vector microphone. Ideal for peripheral edge placement.
Cyber Capsule
Compact PillClean 170px floating status pill featuring a 5-band dynamic equalizer and status divider for distraction-free taskbar integration.
Precision Blade
Studio HUDAudio workstation HUD featuring an 8-segment LED volume meter, hardware acceleration badge, and live decibel telemetry.
Acoustic Ribbon
Harmonic WaveFluid organic harmonic wave with a jewel status indicator, floating suspended smoothly above your active typing area.
Engineering-grade control over local neural compute.
Inspect on-disk model cache sizes, pre-download quantized weights for air-gapped field work, and calibrate microphone dynamics directly on-device.
Quantized Model Library
Pre-download Large v3 Turbo (~1.5 GB), Small, Base, or Tiny models for 100% offline autonomy before boarding flights or entering secure intranets.
SSD Storage Telemetry
Real-time SSD consumption HUD. Reclaim gigabytes of drive space with safe single-click model cache deletion and available capacity telemetry.
24-Segment LED VU Meter
Studio-grade audio level calibration with live microphone testing, precise dB readouts, and pre-amp Whisper Mode voice boost (1.0x–5.0x gain).
Silero VAD & 400ms Pre-Roll
Continuous audio ring buffer preserves syllables spoken fractionally before the hotkey is engaged. Neural VAD trims dead air to cut inference time in half.
Technical Vocabulary Biasing
Feed specialized technical terminology, acronyms, and personal names directly into Whisper's prompt context to permanently eliminate phonetic errors.
Air-Gapped SQLite Scratchpad
Encrypted on-disk dictation history with full-text search, live productivity ROI metrics (words spoken, time saved), and instant SRT or text export.
Local hardware vs. cloud dictation APIs.
Cloud speech services introduce network roundtrip latency, ongoing subscription costs, and sensitive audio leakage. Compare the architectural specifications.
| Architecture Capability | WhisperType (Local Tensor Cores) | Cloud Dictation APIs | Windows Voice Typing (Win+H) |
|---|---|---|---|
| Audio Data Sovereignty | 100% On-Device (Air-Gapped) | Streamed to Remote Servers | Streamed to Remote Servers |
| Offline Operation | Full Functionality (Zero Internet) | Fails Without Active Connection | Severely Degrades or Halts |
| Recurring Cost | $0 Perpetual (Free & Open) | $20 / mo or $0.006 / minute | Included with Windows license |
| Model Architecture | Whisper Large v3 Turbo (809M) | Proprietary Streaming API | Legacy Windows Speech Core |
| End-to-End Latency | ~100 ms (Local Memory Bus) | 400–1400 ms (WAN Roundtrip) | Variable & Inconsistent |
| Active Cursor Focus | Non-Activating Layered Window | Varies by Client Application | Frequently Steals Active Focus |
| Pre-Roll Audio Ring Buffer | 400 ms Pre-Roll Buffer | Cuts Initial Consonants | Clips Early Syllables |
System requirements & quickstart.
WhisperType installs as a self-contained desktop runtime. No external Python installations or command-line configurations are required.
System Requirements
- Operating System: Windows 10 or Windows 11 (64-bit architecture).
- GPU Acceleration (Recommended): NVIDIA GeForce GTX 1060+ or RTX series with 2 GB+ VRAM. Bundled with CUDA 12 & cuDNN 9 runtime libraries.
- CPU Fallback: Modern x86-64 CPU with AVX2 instruction support. Automatically falls back to quantized INT8 inference.
- Storage Footprint: ~2 GB free disk space (includes pre-packaged Whisper Large v3 Turbo model weights).
How to Dictate
- 1. Install: Execute
WhisperType-1.0.0-setup.exe. Setup configures GPU accelerators and system tray integration automatically. - 2. Hold to Dictate: Hold Right Ctrl anywhere in Windows (Word, Slack, Visual Studio, browsers) and speak naturally.
- 3. Release to Insert: Release the key. Transcribed text streams directly into your focused cursor position at 160+ WPM.
- Hands-Free Latch Mode: Quick-tap (< 250 ms) to lock continuous recording mode on or off.
Get-FileHash .\WhisperType-1.0.0-setup.exe -Algorithm SHA256