What is the best open-source meeting transcription tool?

There is no single best tool, because the right one depends on your operating system, your technical skill and whether you need speaker labels. For most people, the starting point is a tool built on OpenAI's Whisper, whose code and model weights are MIT licensed ¹. Meetily is an open-source app built for meetings, and it captures system audio as well as the microphone. Open source means you can read, change and run the code. It does not mean the tool is private, accurate or free to use commercially, so check those things separately.

Whisper is the engine most open-source tools use

OpenAI released Whisper's code and model weights under the MIT License ¹. Tools built on Whisper, such as whisper.cpp and WhisperX, depend on its engine, so the license of the engine matters as much as the license of the app. whisper.cpp is a C/C++ port of the model, and its code is MIT licensed ². It is the lightweight route for a technical user. Our guide to open-source speech recognition explains the setup.

WhisperX adds speaker labels, and its code is BSD licensed

WhisperX adds word-level timestamps and speaker diarization to its output ³. Speaker diarization is the task of working out who spoke when. Its repository is licensed under the BSD 2-Clause License ⁴. The speaker models are downloaded separately, and each has its own terms, so read them before you use WhisperX for a confidential meeting.

For live text, a Whisper stream is the open-source route, and it lags

Live open-source transcription means running a stream, such as the whisper.cpp stream example, on the microphone. It is a starting point rather than a finished meeting product. Expect lag and words that change after they appear. Our question on whether Whisper can transcribe in real time explains why, and our live meeting transcription question compares it with the built-in options.

Meetily is the open-source meeting app to test first

Meetily is released under the MIT License ⁵. Its README says it captures microphone and system audio at the same time ⁶, so it can record a call played on your computer without a bot joining. That makes it the closest match we found to a meeting app. We did not run it on a live call, and the README is the project's own description, so test it with a short meeting before you rely on it.

Meetily also lets you choose the AI provider for summaries. The README lists Ollama, which runs locally, along with cloud services such as Claude and Groq ⁷. Cloud providers are optional, so set the summary option before a confidential meeting. For the comparison with a Mac dictation app, see the best open-source dictation apps for Mac.

Check the model license, not only the app

The license on a tool covers its code, not the model weights it downloads. NVIDIA's Canary 1B model card lists its license as CC-BY-NC-4.0, a Creative Commons license that does not allow commercial use ⁸. If your meeting work is commercial, read the license of each model you choose, and do not assume that an MIT-licensed app comes with a commercially usable model. Is VoiceInk free and open source shows the same split between code and app.

Open source does not change the consent rules

Running a tool on your own machine keeps the audio off a company's servers, and that is a real privacy benefit. It does not remove the duty to tell the people in the meeting that it is being transcribed. The consent steps are the same as for any tool, and they are in our guide to meeting transcription and in transcribing a meeting without a bot.

How to choose

1. Start with your operating system, and check the platforms each project lists in its README.
2. If you need who said what, choose WhisperX, and check the speaker model terms.
3. If you need live text, test a Whisper stream with a short call, and expect revisions.
4. If the meeting is confidential, use a local tool and set the summary provider to a local option, or to none.
5. Check the license of each model you download, and keep a record of the versions you tested.

For the wider choice of tools, see the open-source speech recognition guide and our transcription overview.