Speech to Structured Thought.

Dictation, Context, and AI that works in every app and tab

0 ms
Cloud Latency
Direct local C++ execution
0 Bytes
Data Uploaded
100% on-device privacy
4.2x
Typing Speedup
160 WPM voice vs 40 WPM typing
90+
Languages
Multilingual whisper.cpp engine
THE SUSURRUS SUITE

Capabilities & Architecture

Engineered for speed, privacy, and flawless text execution.

Your Susurrus suite

Dictation Studio

The most productive voice dictation engine ever made

To be radically more productive, speech recognition must strip verbal stutters ("um", "uh"), execute spoken formatting cues, and structure bullet points automatically.

  • 100% On-Device / Offline Speech Recognition
  • Automatic Verbal Stutter & Filler Stripping
  • Direct Cursor Auto-Pasting in any Desktop App
Pastel Sky Background
Real-Time Dictation Studio
RAW AUDIO TRANSCRIPT:

"um so yeah like I will list a few items bold this word that is groceries bananas and milk and email Sarah by five p m..."

↓ Gemma 2B Local LLM Normalization
SUSURRUS PASTED RESULT:

• Groceries: Bananas, Milk, Eggs. Action item: Email Sarah by 5:00 PM.

Context Commands

AI that actually works in every app you use

Highlight text in any desktop application—VS Code, Slack, Outlook, or Notion—hold Ctrl+Shift+T, and speak an instruction to rewrite text in place effortlessly.

  • In-Place Text Editing Across All Desktop Apps
  • Executive Tone Adjustments & Action Item Extraction
  • Zero Clipboard Pollution
Botanical Photography Background
Context Command Sandbox (Ctrl+Shift+T)
SELECTED TEXT:

"Hey team, we might delay the release by two days due to open bugs."

Spoken: "Make this executive and action-oriented."
REPLACED RESULT:

"Team, we are adjusting our release schedule by 48 hours to resolve critical issues."

Dual Local Engine

100% Local. Zero Cloud. Powered by whisper.cpp + Gemma 2B.

Upstream blessed architecture: whisper.cpp handles high-fidelity acoustic speech decoding, llama.cpp handles local context processing.

  • CUDA, Metal & Vulkan GPU Acceleration
  • llama.cpp HTTP Server for Gemma 2B LLM
  • Python stdlib-only Orchestrator
Sunset Photography Background
Local Dual Engine Architecture
Audio Input
VAD Silence
whisper.cpp
CUDA / Metal
llama.cpp
Gemma 2B
SPEED & SAVINGS CALCULATOR

Calculate Your Time & Cost Recovery

See how much time you save each month by speaking at 160 WPM instead of typing at 40 WPM.

Daily Dictation Volume:4,000 words/day
27.5 Hours
Saved Every Month
$120 / Year
Saved vs Paid Cloud STT APIs
Unlocking Ultimate Flow Editorial Silhouette Photo

Unlocking Ultimate Flow.

When all your tools are designed to put you in flow, you work faster, think clearer, and focus on what matters most. Susurrus brings zero-latency, local AI dictation directly to your keyboard—100% offline, private, and free forever.

Loading waitlist count...

Join Susurrus Early Access

Start dictating at 160 WPM across every desktop application with zero cloud latency and 100% on-device privacy.

Primary Use Case:

100% On-Device • Zero Cloud Telemetry • Instant 1-minute setup.