Buzz Transcription Review: Free Whisper Transcription on Windows, Mac and Linux

Sarah Johnson
Writes about field sales, meeting notes and voice-first workflows at ParrotNotes. Every article is reviewed by the ParrotNotes product team before it goes live.

Table of Contents
- 1.Buzz at a glance (as of November 2026)
- 2.What Buzz transcription does
- 3.Buzz transcription on Windows, Mac and Linux: which download to get
- 4.How to transcribe an interview with Buzz
- 5.Speed, accuracy and what your computer needs
- 6.Buzz offline transcription and privacy: what stays on your computer
- 7.Buzz pros and cons
- 8.What users report
- 9.Buzz alternatives
- 10.The verdict of this Buzz transcription review
Leah has a 70-minute interview with a source who asked one thing: don't upload the tape anywhere. She has a Windows laptop, a 9 a.m. deadline and no budget for software. (Leah is a composite, an illustration and not a real reporter.)
That's the job Buzz transcription was built for. The Buzz transcription app is a free, open-source desktop program that runs OpenAI's Whisper speech model on your own computer, so the audio never has to leave it. It works on Windows, Mac and Linux. There's no phone app, and no account or minute limit.
The short verdict: on paper, Buzz is one of the strongest free ways to get an automatic first draft of an interview on a Windows or Linux laptop, and a good free option on an Apple silicon Mac. It asks more of you than a paid app. You pick a model, wait while your computer does the work, and correct the draft yourself.
This review covers what Buzz does as of November 2026, which download fits your computer, how to transcribe an interview step by step, speed and privacy, and when another tool is the better choice.
A note on method: this review is built from Buzz's GitHub repository, its documentation and release files, and its website, all checked on November 21, 2026, plus problems users have reported in public. It isn't based on hands-on testing.
Recording in the field rather than at a desk? ParrotNotes records interviews on your iPhone or Android phone and gives you a transcript and summary when you stop.
Buzz at a glance (as of November 2026)
| Details | |
|---|---|
| What it is | A desktop app that transcribes and translates audio with Whisper and similar models |
| Price | Free, open source (MIT license) |
| Runs on | Windows, Mac (Apple silicon; Intel Macs up to version 1.4.5), Linux |
| Phone app | None from the Buzz project |
| Internet needed? | To download the app and models (and for YouTube links); local transcription runs offline |
| Input | Audio and video files, YouTube links, live microphone |
| Export | TXT, SRT, VTT (plus JSON from the transcript viewer; DOCX with a plugin) |
| Speaker labels | Yes, since version 1.4.1 (not on Intel Macs) |
| Latest release | 1.4.5, August 23, 2026 |
| Who makes it | Created by Chidi Williams; most current work is by maintainer Raivis Dejus |
Sources: the Buzz GitHub repository, the Buzz documentation and its release history, checked November 21, 2026.
What Buzz transcription does
Buzz describes itself as a way to "transcribe and translate audio offline on your personal computer." In practice it does four jobs.
Transcribe files and YouTube links
You import an audio or video file, or paste a YouTube link, pick a language and a model, and Buzz writes a timestamped transcript. You can import a whole folder, and a watch folder can transcribe new files as they land. A command-line version suits anyone who batch-processes recordings.
Live transcription from the microphone
Buzz can transcribe as you speak, with a presentation window for showing captions at an event. The docs are honest that this is "resource-intensive" and may not keep up in real time. Use a smaller model if the text lags. To transcribe audio playing on the computer itself, such as a video call, you need a virtual audio device (BlackHole on Mac, VB-CABLE on Windows). Buzz doesn't capture system sound on its own.
Speaker identification
Since version 1.4.1 (January 2026), Buzz can label who's speaking, either while it transcribes or afterwards from the transcript view. You can leave the speaker count on Auto or set it, and rename "Speaker 0" to a real name. It's not available on Intel Macs.
Translation
The built-in task is Translate to English, which Whisper does itself. Translating into other languages needs an API key for an OpenAI-compatible service, either a cloud one or a model you run yourself. Buzz's docs put the cost at about $1.00 per hour of audio with ChatGPT or Claude models. That part isn't offline.
Editing and export
The transcript viewer has search, playback from 0.5x to 2x, a loop for replaying one line, and keyboard shortcuts. Export gives you TXT for a clean read, or SRT and VTT for captions with timestamps. Plugins added in 1.4.5 include Export to DOCX and an AI Summary that again needs an API key.
Buzz transcription on Windows, Mac and Linux: which download to get
This is where most people get stuck, so here's the decision in one table. It's built from the README, the installation docs, the FAQ and the 1.4.5 release files.
| Your computer | What to install | Engine to pick | Model to start with |
|---|---|---|---|
| Windows PC with an Nvidia graphics card | The 1.4.5 installer: download the .exe and both .bin files, then run the .exe | Faster Whisper (the FAQ suggests an Nvidia card with 6 GB+ of video memory) | Turbo, or Large-V2 for hard audio |
| Windows PC without an Nvidia card | Same installer | Whisper.cpp (uses most graphics chips, or the CPU) | Small, then try a larger one |
| Windows on ARM (Snapdragon X) | Same installer; there's no native ARM build yet | Whisper.cpp | Small |
| Mac with Apple silicon (M1 or later) | The ARM64 .dmg | Whisper.cpp (the FAQ calls it the best option on Macs) | Turbo |
| Intel Mac | The X64 .dmg of version 1.4.5, the last for Intel | Whisper.cpp | Small; no speaker labels |
| Linux | Flathub, Snap or the AppImage | Whisper.cpp, or Faster Whisper with an Nvidia card | Small or Turbo |
Three things to know before you click download:
- The Windows installer is big and unsigned. The 1.4.5 Windows files add up to about 2.7 GB. Windows warns you because the app isn't signed; the docs say to choose More info, then Run anyway.
- Your CPU needs AVX2. The FAQ says very old computers aren't supported.
- Whisper model sizes matter more than the app. Our guide to Whisper transcription apps has a table of the six Whisper sizes and the memory each needs.
Three apps called Buzz: which is which?
Search results mix up three different things, and two of them are out of date.
- Buzz (the one this review covers) is the free, open-source desktop app on GitHub and SourceForge. The website buzzcaptions.com calls it "Buzz Classic."
- Buzz for Mac in the Mac App Store was a separate paid edition. It's been discontinued. Its maintainer wrote in February 2026 that "the App store (Mac native) version has been discontinued," and the store link on buzzcaptions.com no longer opens a listing (checked November 21, 2026). Use the GitHub or SourceForge download instead.
- "Buzz Captions & Translation" on the App Store and Google Play is a live-captioning phone app by a different developer. It isn't the Whisper desktop app and has its own paid plans.
How to transcribe an interview with Buzz
These steps follow Buzz's file import guide as of November 2026, with the settings that matter for interviews.
- Open the import window. Choose Import Media File on the File menu, click the + icon, or press Ctrl+O (Cmd+O on a Mac).
- Pick your recording. Any common audio or video file works, including the MP3 or LPCM files from a Sony voice recorder.
- Set the task and language. Choose Transcribe and select the language by hand. The docs warn that automatic detection "may result in unexpected results."
- Pick the engine and model from the table above.
- Add names to the initial prompt. Under Advanced..., type the names and terms Whisper is likely to misspell: your source's name, the street, the company. OpenAI's own prompting guide recommends this.
- Turn on speaker identification if two or more people talk. Set the number of speakers if you know it; Buzz's docs say a known number helps when it would otherwise merge similar voices.
- Click Run. When the status says Completed, double-click the row to open the transcript.
- Check and export. Search for the quote you need, play that line, fix it, then export TXT for your notes or SRT if you need timestamps.
Take Sam, a freelance podcaster, who drops a 40-minute episode with two guests into Buzz on an M2 MacBook Air. He sets English, types both guests' names and the show's sponsor into the prompt, and sets speakers to 3, counting himself. Then he searches the draft for each name and fixes any the model still got wrong. (Sam is a composite, an illustration of the documented settings, not a real test.)
For what comes after the draft, see how to transcribe an interview for the correction pass and quote checks, and transcripts with timestamps if your editor wants a set format.
Speed, accuracy and what your computer needs
Buzz is as fast as your computer. There's no server doing the heavy lifting, so the same file can take minutes on a gaming PC and much longer on a thin laptop. The FAQ's advice when it's slow: use a smaller model, or switch to Whisper.cpp.
Accuracy comes from the model, not from Buzz. Bigger models make fewer mistakes and run slower. The FAQ adds a warning about Large-V3: it's often the most accurate, but it "can hallucinate or invent words that were never in the audio." For quotes you'll publish, check every line against the tape whatever the model.
Speaker identification is the slow part on most machines. In April 2026 a user on an M3 MacBook Air reported on GitHub that a 45-minute file took 6 to 9 minutes to transcribe but hours to label speakers. The maintainer replied that the feature is "geared towards Nvidia CUDA video cards" and slower on other hardware.
Picture Ravi, a researcher with twelve one-hour interviews on a work laptop with no graphics card. Running all of them through a large model with speaker labels could tie up the machine for a very long time. He uses the small model overnight, labels speakers only on the four interviews he'll quote, and checks those by hand. (Ravi is a composite, an illustration rather than a real case.)
Buzz offline transcription and privacy: what stays on your computer
With a local model, your audio and transcript stay on your computer. You only need the internet to download the app and the models. The FAQ even explains how to copy models to a computer that's never online.
A few things do reach the internet, and you can control them:
- Usage statistics. Buzz's preferences docs say it "collects basic OS name and architecture usage statistics." Setting
BUZZ_DISABLE_TELEMETRY=trueturns this off. - Update checks. These can be turned off with
BUZZ_DISABLE_UPDATE_CHECK=true(since 1.4.5). - API features. The OpenAI API engine, AI translation into other languages and the AI Summary plugin send text or audio to whichever service you configure. Leave them off for sensitive tape.
Recording the interview is a separate question from transcribing it. Consent rules differ by state and country; our guide to one-party consent states is general information, not legal advice. Tell the person you're recording.
Buzz pros and cons
Pros
- Free and open source, with no account, minute limit or watermark
- Runs offline, so the audio stays on your computer
- Windows, Mac and Linux, including GPU support for Nvidia, Apple silicon and most other graphics chips
- Speaker labels, live transcription and translation to English built in
- TXT, SRT and VTT export, folder watch and a command line for batches
- Actively maintained: 1.4.5 shipped in August 2026 and code changes were still landing in November 2026
Cons
- No phone app; you need the recording on your computer first
- Slow on older laptops, and speaker labels are slow without an Nvidia card
- Unsigned Windows installer of about 2.7 GB
- Intel Macs stop at version 1.4.5, without speaker labels
- AI summaries and translation into other languages need your own API key
- You choose engines and models yourself, which takes some trial and error
What users report
Public reviews are thin. SourceForge shows 4.7 out of 5 from three reviews as of November 2026. One reviewer called it "excellent freeware for transcription, translation, and subtitle generation," and another wished the models downloaded to the install folder instead of the system drive.
The GitHub issue tracker is more useful, and it's well kept: 16 open issues on November 21, 2026, most of them feature requests. The ones a journalist would notice:
- Speaker labels are slow on Macs, as above.
- No native Windows on ARM build. A request opened in October 2026 is in progress.
- The installer isn't signed. A request to sign Windows builds is open.
- Confusion over the Mac App Store edition, which users still find linked from the website.
Buzz alternatives
If Buzz doesn't fit, here's where to look. Each link covers the option in full.
- Record and transcribe on your phone. ParrotNotes records in-person interviews on iPhone, Android and Mac and gives you a transcript, summary and search across your recordings. The Free plan includes 100 minutes of recording a month, up to 30 minutes per recording, an AI summary on every recording and 3 audio or video imports a month. Pro is $19.99 a month ($14.99 a month billed annually) for 3,000 minutes and recordings up to 3 hours, with speaker identification and transcription in 99+ languages. Unlike Buzz, ParrotNotes transcribes online, so it doesn't suit tape that must never be uploaded.
- Other Whisper apps. Whisper transcription apps compares MacWhisper, Aiko, Whisper Desktop and others by computer and phone.
- Type it yourself, privately. oTranscribe is a free browser player and editor for manual transcription.
- One short file, no install. Our free audio to text converter transcribes a file up to 10 minutes in your browser.
- Compare the field. Transcription software for journalists and the best free transcription app rank the options by job.
Interviewing away from your desk? Try ParrotNotes free on your next one and have the transcript on your phone before you're back at your laptop.
The verdict of this Buzz transcription review
Buzz gives you what used to cost money: automatic transcription with speaker labels and caption export, on your own computer, for free. It's actively maintained as of November 2026 and runs on Windows, Mac and Linux.
The price is your time and your hardware. Pick the right download and model from the table above, set the language and names by hand, and check every quote against the tape. For sensitive interviews on a laptop, it's the first tool to try.
Your next step: check whether your computer has an Nvidia card or Apple silicon, download the matching Buzz build, and run your shortest recording first.
Frequently Asked Questions
Is Buzz transcription free?
Yes. As of November 2026 the Buzz desktop app is free and open source under the MIT license, with no account, plan or minute limit. Only optional features that call an outside AI service, such as translation into languages other than English or AI summaries, cost money through your own API key.
Does Buzz work offline?
Yes. Once the app and a model are downloaded, Buzz transcribes without an internet connection, and your audio stays on your computer. Its FAQ explains how to copy models to a computer that's never online.
Does Buzz work on Mac?
Yes, on Apple silicon Macs with the ARM64 download. Version 1.4.5 is the last one for Intel Macs, and speaker identification doesn't work on Intel. The separate Mac App Store edition has been discontinued.
Can Buzz tell speakers apart?
Yes, since version 1.4.1. Turn on speaker identification when you import a file, or run it from the transcript view, and rename each speaker. It's much faster with an Nvidia graphics card.
Is there a Buzz app for iPhone or Android?
Not from the Buzz project. The phone app called "Buzz Captions & Translation" is a different developer's live-captioning app. To record and transcribe interviews on a phone, use a phone app such as ParrotNotes.
Related Posts

Sony Voice Recorder: Which Model to Buy, and When a Phone App Is Enough
Sony voice recorder guide: PX470, UX570 and TX660 compared from Sony's own specs, which one to buy, and when a free phone app that transcribes is enough.

Otter AI Chrome Extension: How It Works, What It Records and the Limits
How the Otter AI Chrome extension works in Google Meet and Zoom, what it records, which plans include it and where it stops. Steps checked November 2026.

Noty AI Review: What the Meeting Note Taker Does and What It Costs
Noty AI review, checked Nov 2026: the Chrome extension for Google Meet and Zoom with no bot, Pro at $10 a month, and its limits. Try ParrotNotes free.