声
SYS.ONLINE
MEM: 12MB
KOE声

> INITIATING LOCAL VAD PROTOCOL...

> BYOK DETECTED. ALL SYSTEMS FREE.

YOUR
VOICE
EVERYWHERE.

Lightning-fast voice dictation for Windows. Local VAD. Powered by Groq. Zero subscriptions.

EXECUTE DOWNLOAD
[WARN] REQUIRES WINDOWS 10/11[INFO] REQUIRES GROQ API KEY
NO SUBSCRIPTIONS // LOCAL VAD // OPEN SOURCE // HIGH SPEED // PERFECT PRIVACY // NO SUBSCRIPTIONS // LOCAL VAD // OPEN SOURCE // HIGH SPEED // PERFECT PRIVACY // NO SUBSCRIPTIONS // LOCAL VAD // OPEN SOURCE // HIGH SPEED // PERFECT PRIVACY //

LIVE TRANSCRIPTION DEMO

0/10 used today
Press Start Recording, speak, then Stop Recording to transcribe.
Record speech, then stop to transcribe.

SYSTEM ADVANTAGES

01

LIGHTNING FAST

216x real-time transcription. It's basically instant. Powered by optimized Whisper models on Groq's LPU architecture.

02

PRIVACY FIRST. ALWAYS.

Local Voice Activity Detection processing. Your microphone stays local. Your voice never leaves your boundary until you explicitly trigger dictation.

03

COMPLETELY FREE

No walled gardens. No recurring billing. Bring your own API key and operate indefinitely without limits.

04

NEVER LOSE A THOUGHT

Automatic history saves every transcription. Copy, retry, or delete anytime. Your complete dictation log, locally stored.

[GLOBAL HOTKEY]

THE SHORTCUT TO DOMINANCE.

Configure any binding. Execute it anywhere—VS Code, Chrome, Notion. Focus drops, microphone captures, transcription inserts instantly at your cursor.

COMPARE.

THE CHOICES ARE CLEAR. UPGRADE YOUR WORKFLOW.

METRIC
KOE [V1.0]
A OS NATIVE
COST STRUCTURE
$0 FOREVER
$0
ENGINE TYPE
WHISPER (HIGH)
NATIVE (LOW)
SOURCE
OPEN (GITHUB)
CLOSED

SYS.FAQ

[Q 01] IS KOE REALLY FREE?
> RESPONSE:Yes, completely free and open source. No hidden costs or subscriptions to be found in the core logic.
[Q 02] DO I NEED AN API KEY?
> RESPONSE:Yes. Koe operates on a BYOK (Bring Your Own Key) model. Acquire a free hardware-accelerated key from Groq to activate full capabilities.
[Q 03] IS MY AUDIO MINED FOR DATA?
> RESPONSE:Negative. VAD is processed 100% locally. Audio transmission only occurs explicitly on hotkey trigger directly to the AI provider. No telemetry.