Always ready.
Never cloud-dependent.
Wake-word detection, transcription, routing, and fallback reasoning all run on your Windows machine.
VOICEPRIVATEOFFLINE-FIRST
Local system online WIN 10/11
VOICE → INTENT → ACTION
Your voice.
Your machine.
A hands-free Windows command layer that handles everyday requests instantly, then brings in local AI only when language gets ambiguous.
233stress-tested phrases
<LLMfast path comes first
LOCALspeech + reasoning
JARVIS CORE / 03SIGNAL 98%
STANDBY
>
Say a command or choose a route below.
ROUTER / READY00.0 MS
OPENWAKEWORDFASTER-WHISPEROLLAMAPLAYWRIGHTPYTHON 3.11+OPENWAKEWORDFASTER-WHISPER
01 THE SYSTEM
Intelligence that knows
when not to think.
Most voice assistants send every utterance through a model. Jarvis treats latency as a design constraint: common commands stay deterministic; context and reasoning are earned only when needed.
Always ready.
Media that listens
Search music, control playback, volume, tabs, and fullscreen without breaking focus.
07media actions
Workflows, not shortcuts
Launch multi-step study, assignment, bug bounty, or music modes with one phrase.
04bundled modes
COMMAND PREVIEW
you open my research setup
jarvis bug bounty workflow active
✓ Firefox workspace
✓ Notes and proxy
✓ Target scope
✓ Notes and proxy
✓ Target scope
Power, with a brake pedal.
High-impact system actions require spoken confirmation. Allowed apps are explicit. Tool schemas narrow as certainty improves.
02 ROUTING ENGINE
Fast by default.
Smart on demand.
Each utterance takes the shortest trustworthy route to execution.
00
~0ms
Exact command
Known phrasing maps directly to an approved action.
pause · volume · open · workflow01
FAST
Context repair
Vocabulary and entity data clean up imperfect speech.
artists · titles · URLs · apps02
LOCAL
Heuristic route
Intent patterns resolve common ambiguous forms.
questions · search · navigation03
QWEN
Reason + act
The local model receives only the narrowed tool set.
low-confidence commands only03 BRING IT ONLINE
Four moves.
One voice.
Built for Windows 10/11, Python 3.11+, Ollama, and a microphone. GPU acceleration is optional.
Read full installation guide ↗- 01Clone the source
git clone .../Jarvis-Voice-Assistant - 02Install the voice stack
pip install -r requirements.txt - 03Configure your devices
python scripts/list_mics.py - 04Say the words
.\run_jarvis.ps1