Meetily: Private AI Meeting Notes, 100% On Device

0:00 This is mely and it fixes something that should genuinely worry you about every other AI noteaker. It records, transcribes, and summarizes your meetings 100% on your own device. No audio, no transcripts. Nothing gets uploaded to somebody else's cloud. It's open source. It's free to self-host and your private conversations finally stay private. In a world where every app wants a copy of everything you say, that is a genuinely radical idea and Meily makes it the default. Meily is built by a team called Zachariah Solutions whose whole mission is powerful AI that respects your data sovereignty. It's MIT licensed written in Rust and it has already earned over 16,000 stars on GitHub. That Rust foundation matters because it's what lets the transcription run fast enough to keep up with a live conversation without melting your laptop. And being MIT licensed means you can read every line, fork it, and audit exactly what it does with your audio, which for a privacy tool is really the whole point. And stick around because there's one design decision underneath me that makes it genuinely private, not just privately marketed. And I'll show you exactly what it is. So, what's the actual problem here? Think about the tools most people use. You invite a cloudbot into your call and now a recording of your most sensitive conversation, salaries, strategy, patient details, legal matters, lives on a server you do not control with data practices you probably never read. For a lawyer, a doctor, or anyone under compliance rules, that is inconvenient.

1:32 It's a liability. Medially flips the model completely. The recording, the transcript, and the summary all stay on your machine. Full stop. We've all clicked accept and watched a little bot join the call to take notes. It feels helpful right up until you remember that a full recording of the whole thing now sits in a vendor's database indexed and possibly feeding the next model. Once that audio leaves your device, you have effectively lost control of it forever.

1:58 And this clearly resonates with people. According to Midi's own site, more than 297,000 users have already downloaded it. on top of those 16,000 GitHub stars. For a privacy tool in a space crowded with slick cloud products, that kind of adoption tells you a lot of professionals were quietly desperate for an option that simply doesn't phone home. Those are two different numbers measuring the same thing. Developers starring the code and regular users actually installing the app and both are pointing sharply up and to the right.

2:28 Here's what it actually does. medially sits on your desktop and listens to any meeting whether that's Zoom, Teams or Google Meet. It transcribes everything live in real time as people speak and it can tell the speakers apart. When the call wraps up, it generates a clean summary with the key points and the action items. And that whole capture, transcribe, and summarize loop happens right there on your laptop, fully offline if you want it to be. You can search back through past meetings, copy the transcript straight out, and import recordings you already have. Nothing about the workflow feels like a compromise you're making for privacy. It just works like a normal polished notetaker, except the entire brain of it lives on your own computer instead of in a data center. The feature list reads like a privacy wish list. Everything is local first, so nothing leaves your computer. You get live transcription and AI summaries. It runs on Mac OS, Windows, and Linux. And crucially, you choose the AI brain. Point it at Olma running locally or plug-in claude gro or any open AI compatible endpoint if you'd rather. It slots straight into the meeting tools you already use with no workflow change required.

3:38 That provider choice is the sleeper feature. If you're fully offline or genuinely paranoid, keep everything on Ola Ma and never touch the internet. If you have a FastCloud key and don't mind using it for just the summary, you can do that, too. You decide exactly how much, if anything, ever leaves your machine. And here's that design decision I promised. The thing that makes this genuinely private. Meily runs the AI models themselves on your hardware. The transcription engine, whisper or parakeet, and a local model through Ola for the summary. All execute on your machine. That means you can literally pull your network cable, run an entire confidential meeting, and get a full transcript and summary with exactly zero bytes leaving the building. That's privacy you can verify. Not just privacy you have to trust. Most privacy tools ask you to take their word for it.

4:24 Medially lets you prove it yourself. Open a network monitor. Start a meeting and watch the traffic stay flat at zero. No telemetry quietly calling home. No analytics beacon. No hidden upload. What runs on your machine stays on your machine. And you can confirm that in about 30 seconds. And it's a serious healthy project, not a weekend experiment. over 16,000 stars, more than 1700 forks, 11 releases shipped, and it's all MIT licensed. It's built on Rust for Speed. The live transcription is claimed to run around four times faster than you'd expect. And the community around it is active and clearly growing. 11 releases in around 18 months is a steady, sustainable pace, and issues are actually getting triaged and closed, which is exactly the health signal you want before you build a workflow on top of something. Now, the honest context because Meily has a commercial side. The open source community edition is the free self-host able core we've been talking about and it is genuinely capable, but there's also a paid melee pro with extra accuracy, advanced exports, and team features. And the website leads pretty hard with buy now and a free trial.

5:30 That's a fair open core model. Just no going into some of the Polish lives behind the payw wall. And running the models locally does ask something of your hardware. So, a very old laptop may struggle with the heavier transcription. None of that makes it a bad deal, though. Plenty of great open source projects fund themselves with a paid tier, and honestly, that's often how a project like this stays alive and maintained. The important part is that the free self-hosted core genuinely covers the main job, private local meeting notes without paying a scent.

6:03 So, if you take meetings that actually matter and you'd rather not hand them to a cloud you can't see into, Meily is absolutely worth a look. Grab the open source build, point it at a local model, and keep your conversations yours. Links to the site and the GitHub repo are both in the description. If you like open source that puts your privacy first, subscribe because I find one of these every single week. I'll catch you in the next