skip to content
Andrew Marder

Introducing whispy: Easy Speech-to-Text on Mac

/ 2 min read

Table of Contents
View more blogs with the tag tools

If you’re a writer who tends to over-edit while drafting (like me), or you simply want an easy way to convert speech to text on your Mac, I’ve built a tool that might help: whispy.

What is whispy?

whispy is a Python wrapper around whisper.cpp that makes speech-to-text transcription incredibly simple. whispy does three things:

  1. Record audio from your microphone
  2. Transcribe that audio to text using OpenAI’s Whisper models
  3. Automatically copy the text to your clipboard to paste anywhere

Motivation

I have a problem: I edit way too much while writing. It slows down my writing process because I constantly second-guess myself and revise sentences before I’ve even finished my thoughts.

Speech-to-text lets me bypass this tendency entirely. I can just speak my ideas fluently without the temptation to edit, then clean up the transcription afterward. The result is faster first drafts and better flow.

Technical Foundation

whispy is built on top of whisper.cpp, which is a highly optimized C++ implementation of OpenAI’s Whisper speech recognition model. On Apple Silicon Macs, whisper.cpp leverages the Accelerate framework and Metal for incredible performance - often transcribing faster than real-time even with high-accuracy models.

The problem? whisper.cpp can be frustrating to install and set up. You need to clone repositories, run build scripts, download models manually, and figure out command-line arguments.

whispy solves this by handling all the setup automatically:

  • Auto-installation: Clones and builds whisper.cpp for you
  • Model management: Downloads models as needed
  • Simple interface: Just run whispy record and start talking

See It In Action

Here’s a one-minute demo showing how whispy works:

Try It Out

Installation is straightforward:

Terminal window
pip install git+https://github.com/amarder/whispy.git

This automatically sets up everything you need, including building whisper.cpp with GPU acceleration.

Basic Usage:

Terminal window
# Record and transcribe in one command
whispy record

If you run into any issues, please open an issue on GitHub - I’ll help you get set up.

Whether you’re a writer looking to improve your drafting process, a researcher conducting interviews, or anyone who needs reliable speech-to-text on Mac, whispy might be a great tool for you.