Are you looking for Mac software to transcribe audio files into written text files? There are plenty of paid tools, like Otter.ai and Trint, that can cost real money when you start using them too much. But MacWhisper is the free solution the internet has been waiting for. Or, at least, the one I’ve been waiting for.
OpenAI released an incredibly powerful free program back in September 2022 that allowed people with some tech savvy to convert their audio and video files into incredibly accurate transcripts. The problem was that it wasn’t very accessible to the average user, since you needed to be familiar with using commands in the Terminal app.
But with MacWhisper, which was first spotted by 9to5Mac, someone has finally developed a GUI, or graphical user interface, that builds on top of OpenAI’s Whisper and leaves all the confusing tech stuff under the hood, letting you just drag and drop your audio file and get a transcription very quickly. And MacWhisper is extremely accurate because OpenAI’s Whisper has been trained on 680,000 hours of audio data on the web.
One of the nice things about MacWhisper is that you’re not uploading the audio or video file that you want transcribed to the internet. That’s important for journalists who may be working on a sensitive story, or anyone who’s just generally wary of uploading their personal files to the cloud. The file stays on your desktop, meaning that there’s no chance anyone could intercept your audio.
Again, MacWhisper is just a new user interface for a free program that was introduced by OpenAI late last year, but it’s really going to make audio transcription easy for just about anyone. Because I’ve been using Whisper and have found it tremendously useful since it was released. And even I will now be using MacWhsiper, because it’s sure to speed up my workflow.
MORE FOR YOU
As just one example, after news of crypto company FTX’s collapse, I used Whisper to transcribe dozens of hours of old YouTube interviews with company founder Sam Bankman-Fried. From there, I was able to search the transcripts and find weird and interesting ideas for articles to write, like when SBF said in June 2021 that blockchain would’ve fixed Enron. SBF’s comparison to Enron was an odd coincidence, given everything we now know about how FTX collapsed. But I never would’ve found that moment in an old obscure interview without having dozens of transcripts I could search. And that process, while easier than listening to hours and hours of tapes, was still clunky and not very automated.
After getting Whisper installed on my machine, I used to type “whisper audiofilename.mp3 —model tiny.en” into Mac’s Terminal app any time I wanted to transcribe a file. With WhisperMac, I can just drag and drop my audio file and it does its magic without any extra typing.
I tested out MacWhisper with an episode of NBC Nightly News that I downloaded from YouTube. And aside from a few errors, like interpreting the word “windchills” as “windshills,” and hearing Lester Holt as “Lester Hold,” the transcript was very accurate. MacWhisper is free, but there is a paid tier of the software that could be slightly more accurate because it uses OpenAI’s more intensive training models, but the free version will work just great for most people.
MacWhisper allows you to export the transcribed words in a few different formats, including plain text, CSV or even a subtitles file format like SRT and VTT.
And MacWhisper isn’t just for English. You can transcribe audio in 100 different languages, which is incredibly handy. The only feature MacWhisper doesn’t have, which I’ve enjoyed playing around with on Whisper in the Terminal app, is language translation. I’ve translated viral videos in Russian and Chinese and have been told by people who speak the languages on Twitter that the translations are accurate.
The other feature that MacWhisper doesn’t have, though it’s also not available from OpenAI’s Whisper, is the ability to differentiate speakers. For example, if you’re transcribing an interview with two speakers, there’s no clear markings that show when one speaker has stopped and another has started. But I suspect someone will build that capability eventually. Paid services like Trint do this automatically, and it would be great to have for free eventually, but you do get what you pay for.
There’s also a feature with MacWhisper where you can convert speech from your computer’s microphone into text, which could be handy if you want to transcribe something on the fly.
I spent months wondering when someone would finally develop a GUI for Whisper, as a few people gave it a shot but didn’t create anything very user friendly. But MacWhisper, which was developed by Jordi Bruin, is the software I’ve been waiting for. Bruin has also developed software like MacGPT, which allows users to access ChatGPT from their menubar.
You can download MacWhisper for free, with email registration, at Bruin’s website.