Qualitative researchers often rely on software products to transcribe and translate interview data. The sensitivity of these data requires researchers to handle them locally. Running Whisper locally will help researchers adhere to the best IRB practice without relying on software products. This article shows how to run different sizes of Whisper locally on Mac fast. If you have never opened the terminal on your Mac before, I got you! If you frequently visit GitHub, this link is for you. If you want to use a software product that builds on top of Whisper, check out this link. I have not tried this product and am not affiliated with it in any way.

Prepare your Mac

First, let’s open your terminal and install Homebrew. You can open the spotlight search on your Mac and type in terminal. Homebrew is the package manager for macOS. We need Homebrew to install FFmpeg, which handles audio and video files. Whisper accepts multiple audio formats, but we need .wav format to use the community solution offering a speed bump. We also need Homebrew to install Git, a source code management tool. GitHub is built on top of Git.

Install Homebrew

Copy and paste this line on your terminal. You can find this code on Homebrew homepage.

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Type “brew” to check if you have installed Homebrew successfully.

Install FFmpeg with Homebrew

Copy and paste the following code. You can also find more information on installing FFmpeg with Homebrew here.

Install Git with Homebrew

It follows the same “brew install” formula. Copy and paste the following code. You can also find more information on installing Git with Homebrew here.

brew install git

Install cmake with Homebrew

cmake is a tool for building C++ code.

brew install cmake

Now, we are ready to get started by cloning this GitHub repository. It is a port of OpenAI's Whisper model in C/C++, which offers a speed boost. Clone means having a copy of the repository hosted on Github locally on your laptop.

Get Whisper

Let’s find a location to save the repository. First, let’s check what your current working directory is by running this code:

pwd

Use “cd” command to navigate to a different directory. Use “..” to navigate back to the parent directory. If you are in a subdirectory under Documents, running the following code will navigate you to Documents. You can find more information on navigating files using a Mac terminal here. Or ask ChatGPT.

Clone

Now, let’s clone the repository.

git clone https://github.com/ggerganov/whisper.cpp

cd whisper.cpp 

Prepare your audio

Convert your .mp3 file to a .wav file. If your audio files are in other formats, simply replace the mp3 file with your file in other formats.

ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le output.wav

Once you have the .wav file, drag and drop it to the samples folder under the whisper.cpp folder.

We are ready to run Whisper now!

Run Whisper

The base or tiny model will work fine if your task is transcribing texts in English.

If your task is more complicated, try experimenting with different model sizes to find the model size that fits your use case!

The following code shows how to run the base Whisper model.

First, download one of the Whisper models converted in ggml format. For example:

sh ./models/download-ggml-model.sh base.en

Now build the whisper-cli example and transcribe an audio file like this:

# build the project
cmake -B build
cmake --build build --config Release

# transcribe an audio file
./build/bin/whisper-cli -f samples/jfk.wav

For a quick demo, simply run make base.en.

The command downloads the base.en model converted to custom ggml format and runs the inference on all .wav samples in the folder samples.

For detailed usage instructions, run:

./build/bin/whisper-cli -h

UCR Library relaunches its Digital Scholarship services with one-on-one or group consultations and monthly workshops and meetups!

DS workshops cover topics about integrating digital tools, methods, and technologies into their research, teaching, learning, and publishing.

DS meetups invite guest speakers from UCR and beyond and cultivate community conversations on digital tools, methods, and technologies and facilitate collaborative learning and connection across all disciplines at UCR.

Our guest speakers for the Spring DS Meetup Series include:

Dr. Covadonga Lamar Prieto, the Interim Associate Dean for Student Academic Affairs in CHASS and an Associate Professor in the Dept. of Hispanic Studies. She also runs Spanish of California Lab.

Dr. Anthony Russell Jerry, an Associate Professor at the Department of Black Study. His DS projects include the Cultural Media Archive, the Learning Archive, and the Empathy Archive.

Dr. Catherine Gudis, Teresa and Byron Pollitt Endowed Term Chair for Interdisciplinary Research and Learning in the Humanities & Social Sciences at UCR. Her work includes A People’s History of the I.E. and her research has a focus on digital humanities in Southern California in general.

Subscribe to stay updated on UCR DS events!

Subscribe now

DS Events Beyond UCR

Soo many goods at Digital Humanities Summer Institute! Courses we are eyeing include:

DH Sample Platter

NLP, LLMs, and Network Science Apps for Text and Media Analysis and Creativity

Coding Fundamentals for Humanists

Dream Lab and Digital Humanities at Oxford Summer School have very interesting course offerings too!

Tips

As the Digital Scholarship Resources website shows, many many commercial and open-source tools are available for DS projects. How do we know which one to use? How can we have time to learn to use all these tools?

Those are indeed very important questions, and, hopefully you will develop your own answers as you spend more time on this newsletter!

But for now, I would like to share a tip for exploring no-code or low-code website builders to display your DS projects or portfolio.

Remember to ask these three questions:

Where can I put all my content?

Where can I change the overall style?

Where can I edit the site information?

Also remember the fundamentals of how a browser displays a website:

Browser to Someone Else’s Computer (also known as Server): Where is example.com? I am requesting its files.

Server to Browser: I got your request. I am responding by sending files over!

Browser to Itself: I will keep some files so next time, I can display this page faster!

Hopefully this issue will be helpful in your DS journey!

Subscribe now