Free guide · Whistle (cactus-needle 3.2.0)

Transcribe audio on your own computer with Whistle, in 5 minutes

Whistle is a free speech to text model from Cactus Compute. The whole model is one 16.9 MB file that runs on your own CPU, in seven languages. This guide installs it, transcribes your first file and turns on live transcription from your microphone.

A ready prompt that lets Claude Code or Codex set it up for you
The install and your first transcript in two commands
5 steps with screenshots, from install to live microphone mode
What it costs, where it is weaker than Whisper and fixes for common errors

A look inside

What's inside

The agent prompt

Paste one prompt into Claude Code or Codex. It checks Python, installs Whistle in its own environment and transcribes a test file.

Names it gets right

How to pass your client names and product words, force a language and get a time stamp for every word.

Live from the microphone

The one extra package for microphone input, and the terminal command that writes down what you say as you say it.

🔒 Full guide after signup

Good to know

Questions, answered

Is this guide free?

Yes. You sign up with your email once and get this guide plus every other guide in the library.

Is Whistle free?

Yes. The model and the Python package are open source under the Apache 2.0 license. It runs on your own CPU, so there is no account, no API key and no cost per minute of audio.

Which languages does it understand?

English, German, French, Spanish, Italian, Dutch and Polish. It detects the language on its own, or you set it with language="de" for example. Other languages, like Portuguese, are not supported yet.

What happens after I enter my email?

We send a confirmation email. Click the link and the full guide opens. You also get the ai-toolfinder newsletter every Friday and can unsubscribe anytime.