Buzz is a free, open source desktop app that transcribes and translates audio on your own computer, so recordings never leave your machine. It is worth it for anyone with sensitive audio and a reasonably modern computer, but people who want a quick web app with no setup will find it more work.
What Buzz is
Buzz is a desktop app that transcribes and translates audio offline on your personal computer, using OpenAI's Whisper speech model Buzz GitHub ↗. It is open source under the MIT license, and the project has about 21,900 stars on GitHub Buzz GitHub ↗.
It runs on Mac, Windows and Linux Buzz GitHub ↗. Mac users download a .dmg file and Windows users an installer, both from SourceForge. Linux users can pick Flatpak, Snap or AppImage, and developers can install it from PyPI with Python 3.12 and ffmpeg Buzz GitHub ↗.
Because the work happens on your own hardware, there are no minute caps or monthly limits Buzz GitHub ↗. Hands on testing is in progress; we will update this review with results.
What it does well
Privacy is the main reason to choose Buzz. Audio is processed on your computer instead of being uploaded to a company's server Buzz GitHub ↗. That makes it a strong free choice for client calls, patient notes or legal recordings that should stay in house.
It covers more than plain file transcription. Buzz can transcribe audio and video files, YouTube links and live speech from your microphone, with a presentation window for live events Buzz GitHub ↗. It can separate speech from background noise before transcribing, identify speakers, and export TXT, SRT and VTT files for documents and captions Buzz GitHub ↗.
The transcript viewer has search, playback controls and speed adjustment Buzz GitHub ↗. For heavier use there is a watch folder that transcribes new files automatically, a command line tool and a plugin system that includes AI summaries Buzz GitHub ↗.
You also get a choice of engines. Buzz supports Whisper, Whisper.cpp and Hugging Face models Buzz GitHub ↗. Its FAQ says Whisper.cpp runs on any GPU or CPU and is recommended for Macs, while Faster Whisper is much faster on an Nvidia GPU with at least 6GB of video memory Buzz FAQ ↗. The Hugging Face option adds models such as MMS, which covers over 1,000 languages Buzz FAQ ↗. If speed matters more than privacy for a given file, you can add an OpenAI API key to run recognition on a remote server Buzz FAQ ↗.
GPU acceleration is supported on Nvidia cards through CUDA, on Apple silicon Macs, and through Vulkan on most GPUs including integrated ones Buzz GitHub ↗.
Where it falls short
Setup takes more effort than a web app. You install the program and download a model before you can transcribe. To use it fully offline, you download models on a connected computer and copy them over by hand Buzz FAQ ↗.
Speed depends on your machine. The FAQ says smaller models are faster but less accurate, and larger ones need more hardware or time Buzz FAQ ↗. It also warns that Large-V3, often the most accurate model, can hallucinate words that are not in the audio Buzz FAQ ↗.
Hardware rules can block older computers. Buzz needs a CPU with AVX2 support, so very old computers are not supported Buzz FAQ ↗. Current Mac versions require Apple silicon, and the last version for Intel Macs is 1.4.5 Buzz GitHub ↗.
The Windows installer is unsigned, so Windows shows a security warning and you have to click More info, then Run anyway Buzz GitHub ↗. Some features are also missing with certain engines: translation and word level timings do not work with Hugging Face models Buzz FAQ ↗.
There is no company support desk or cloud sync. Help comes from the documentation and the GitHub project.
Pricing and plans
Buzz is free and open source under the MIT license Buzz GitHub ↗. There are no plans, no subscription and no usage limits, because your own computer does the transcription Buzz GitHub ↗. There is no trial because there is nothing to buy.
The only possible costs are indirect. A faster computer or a GPU will speed things up, and if you choose to connect an OpenAI API key for remote recognition, OpenAI bills that usage separately Buzz FAQ ↗.
Who should (and shouldn't) use it
- Good fit: anyone transcribing sensitive recordings that should not be uploaded.
- Good fit: people with long or frequent recordings who do not want monthly caps.
- Good fit: video makers who need free SRT or VTT caption files.
- Skip it if: you want a web app that works in a browser with no install.
- Skip it if: you have an Intel Mac and want the latest version, or a very old PC without AVX2.
- Skip it if: you want meetings joined and recorded automatically.
Alternatives worth a look
- TurboScribe: web based, free for 3 files a day of up to 30 minutes each, with speaker labels TurboScribe pricing ↗.
- Otter.ai: free for 300 minutes a month, with live transcription of meetings Otter pricing ↗.
- Fathom: free unlimited recording and transcription of video calls Fathom pricing ↗.
- Apple Notes: built into iOS 18 and later, with a live transcript for quick voice notes Popular Science ↗.
Where it appears
Best for privacyThe Best Free AI Transcription ToolsWhat’s good
- Audio never leaves your computer
- No usage caps
- Mac, Windows and Linux
What to know
- More setup than a web app
- Unsigned Windows installer
Key features
- Offline transcription and translation with Whisper
- Files, YouTube links and live microphone
- Speaker identification
- Exports TXT, SRT and VTT
- GPU acceleration (CUDA, Apple silicon, Vulkan)
- Command line, watch folder and plugins
Questions people ask
Is Buzz really free?
Yes. Buzz is open source under the MIT license, with no usage limits, because it runs on your own computer Buzz GitHub ↗.
Does Buzz work offline?
Yes. It transcribes and translates on your computer. For a computer with no internet, you download models elsewhere and copy them over Buzz GitHub ↗ Buzz FAQ ↗.
Does Buzz work on Windows and Intel Macs?
It runs on Windows, with an unsigned installer that triggers a warning. Current Mac versions need Apple silicon; 1.4.5 is the last version for Intel Macs Buzz GitHub ↗.
Which Buzz model should I use?
Smaller models are faster and less accurate. The FAQ recommends Whisper.cpp for Macs and Faster Whisper for Nvidia GPUs with at least 6GB of video memory Buzz FAQ ↗.
Where this comes from
Checked Oct 9, 2026. Vendor pages describe their own products; independent sources add outside views. Hands on testing is in progress.
01