Skip to content

mgelsinger/speakeasy-lite

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Speakeasy Lite

Local Windows dictation utility. Press a button, speak, stop — text appears where your cursor is.

No internet required after the model downloads on first run.


Requirements

  • Windows 11
  • Python 3.11 or 3.12
  • A microphone
  • ~1.5 GB disk space for the Whisper model

Setup

.\scripts\setup.ps1

This creates a .venv, installs all dependencies, and downloads the Whisper Turbo model on first transcription.


Run

.\scripts\run.ps1

A tray icon appears in the system tray (bottom-right).


Usage

Trigger Action
Mouse Button 5 (forward thumb) Toggle recording on/off
Ctrl+Alt+D Toggle recording on/off
Tray icon → Toggle Recording Toggle recording on/off

Flow:

  1. Click into the text field you want to type into (Discord, browser, etc.)
  2. Trigger recording (button or hotkey)
  3. Speak
  4. Trigger again to stop
  5. Text is pasted automatically

Tray icon colors:

  • Green — idle
  • Red — recording
  • Orange — transcribing (Whisper running)

Build (single .exe)

.\scripts\build.ps1

Output: dist\SpeakeasyLite.exe

The first run will still download the Whisper model (~1.5 GB) unless you pre-cache it.


Troubleshooting

No tray icon appears

  • Check speakeasy.log in the project root for errors
  • Make sure .venv is set up: run .\scripts\setup.ps1

Recording doesn't start

  • Check that your microphone is the Windows default recording device
  • speakeasy.log will show the error

Text doesn't paste

  • Transcription runs after you stop recording — wait for the tray icon to turn green
  • Make sure the target text field is still focused when you stop recording
  • Some apps (games, elevated-privilege apps) block simulated input; try a browser instead

Whisper model download fails

  • The model downloads from Hugging Face on first transcription
  • If behind a proxy, set HF_HUB_OFFLINE=1 after a successful download to prevent re-checks

Slow transcription

  • Without a CUDA GPU, Whisper Turbo on CPU takes 5–15 seconds per clip
  • Check speakeasy.log — it logs whether GPU or CPU was used at startup

Ctrl+Alt+D conflicts with another app

  • Edit app\hotkeys.py line with HotKey.parse and change the combination

Log file

speakeasy.log in the project root. Rotates automatically on each launch (appends).

About

Simple voice to text

Resources

Stars

0 stars

Watchers

0 watching

Forks

Packages

 
 
 

Contributors