Free & open source speech-to-text

Speakfreely.

Press a shortcut. Speak naturally. Text appears where your cursor is — transcribed by on‑device AI that never phones home.

Download Echo — it’s free
macOSWindowsLinuxNo account needed

Voice to text, anywhere you type

Echo works system-wide. Trigger it from any app, and your words appear right where your cursor is.

Meeting Notes — Notion

Q4 Planning Session

Maya Chen, Josh Rivera, Li Wei Zhang

The backend migration finished ahead of schedule.

Design team flagged two accessibility issues in the nav.

We agreed to push the launch to March fifteenth...

Your voice never leaves this machine.

Zero bytes sent to any server. Zero accounts required. Zero telemetry. Everything happens on your hardware, in your RAM, under your control.

🎙
Your mic
🧠
Local AI
Aa
Your text
0
bytes uploaded
0
accounts needed
MIT
licensed

A native app that disappears

Configure shortcuts, pick your model, and review every transcription. Built with Rust and Tauri for a fast, minimal footprint.

Schedule a follow-up with the design team next Tuesday...
Echo

Transcription History

Today
2 min ago·8.3s·24 words

We agreed to push the launch to March fifteenth to give the design team two extra weeks for the accessibility audit.

14 min ago·5.1s·15 words

The backend migration finished ahead of schedule. We can start QA testing by Thursday.

28 min ago·4.7s·16 words

Check if the user's session has been revoked in Redis before allowing the request through.

1 hr ago·3.9s·13 words

Hi Sarah, yes we're on track. The frontend should be done by Thursday.

Six engines, one shortcut

From lightning-fast CPU models to high-accuracy GPU options. All run 100% locally — pick one and forget about it.

RecommendedNVIDIA Parakeet478 MB

Parakeet V3

Fast, accurate, auto-detects language

Accuracy80%
Speed85%

All models are downloaded once and run entirely on your device.

Your voice deserves better than a server

Every feature works offline, out of the box. No accounts, no cloud, no compromises.

100% Private

All processing happens on your device. No audio or text ever leaves your machine.

Processing locally

Global Shortcuts

Configure hotkeys and push-to-talk to trigger transcription from any app.

Echo Shortcut
CtrlShiftE
Push to Talk

Voice Activity Detection

Silero-based VAD filters silence, so you only transcribe speech.

Silence filtered
Voice detected

File Transcription

Drag & drop audio or video files for batch transcription with progress tracking.

interview.mp3

24.3 MB

Done

meeting.wav

156.7 MB

100+ Languages

Whisper supports 100+ languages. Parakeet auto-detects for you.

Language
Auto Detect
🇺🇸English🇫🇷French🇩🇪German🇯🇵Japanese🇪🇸Spanish+95 more

AI Post-Processing

Clean up transcriptions with LLM-powered refinement. Fix grammar and filler words.

Raw input

so basically the the team needs to uh finalize the the designs by friday

LLM post-processed

The team needs to finalize the designs by Friday.

LLM Tool Calling

Speak natural commands and Echo routes them to tools via LLM function calling.

Tool Calling

Speak a command, Echo calls the tool

"Send a message to Sarah saying I'll be late"
send_message(to: "Sarah", body: "I'll be late")

Cross Platform

Native apps for macOS, Windows, and Linux. Built with Rust and Tauri.

macOSIntel & Apple Silicon
Windowsx64
LinuxAppImage & .deb

Download Echo — it's free

No account. No tracking. No cloud. Free forever under the MIT license.

macOS

macOS 11.0+

Windows

Windows 10 (64-bit)

Linux

glibc ≥ 2.28

Looking for other versions? View all releases

Common questions