Meetily Review (2026): The Competitor That Agrees With Me

Short answer - should you use Meetily?

• Use it if you're on Windows, or your team is split between Windows and Mac. That's the cleanest reason there is, and no Mac-only tool can argue with it.
• Use it if you want transcription that can't leave the machine by construction. Community has no cloud transcription option in its settings at all.
• Use it if you care about community size and project momentum. 32.6k GitHub stars and close to a hundred commits in the last three months.
• Skip the free edition if you need to know who said what. Community has no speaker labels, and because of how it captures audio it doesn't even have "me vs them".
• Skip it if you want to ask questions across your meeting history. Chat is listed as "Soon" on every Meetily tier.

Bias: I make Humla, which is the most directly competing product to Meetily that exists, and I'd love you to download mine. Every Meetily fact below is from its GitHub repo, its source code, and meetily.ai on 10 November 2026.

#The awkward bit, first

Every other review I've written had an easy spine. Otter is cloud, Granola is cloud, Fireflies emails your client a recap. I can be fair about all of them and still end on "but your audio went to a server," and that sentence does a lot of quiet work for me.

It does nothing here.

Meetily is local. Meetily is open source, MIT. Meetily has no bot. Meetily lets you run summaries through Ollama so nothing leaves the laptop, or bring your own key if you'd rather. Read their README and then read my homepage and you'll find the same paragraph written twice by two different teams in two different countries. It's the one competitor whose thesis is mine, which is great for the thesis and deeply inconvenient for my marketing.

So my Bias - and I want you to picture it as a small, well-dressed man sitting on my shoulder, quietly taking notes on everything Meetily does worse - has a problem. The usual arguments are off the table. He'll have to work for it.

(He's already muttering about the star count. Let's get that out of the way.)

#The numbers, which are not close

From the GitHub API on 10 November 2026:

Meetily's Pro page also claims 739K+ downloads. I can't check that number and I'm not going to pretend I can, but nothing about the GitHub side makes it implausible.

That's roughly a hundred-to-one on stars. And it's not a dead hundred-to-one: the v0.4.1 release notes list eight first-time contributors, fixing things like Bluetooth mics dropping out mid-recording and the voice detector chopping live speech into fragments. That's what a healthy open-source project looks like. People show up, scratch their own itch, leave the code better.

My Bias would like you to know that stars are a vanity metric. My Bias also refreshes my own star count more often than he'd admit, so let's not take his word for it.

#What the free version actually does

Meetily Community, v0.4.1:

• Captures your mic and the system audio together, live.
• Transcribes on-device with NVIDIA's Parakeet (the one it recommends, for speed) or Whisper. I went looking in the source for cloud transcription and found Deepgram, ElevenLabs, Groq and OpenAI in the settings component, all commented out. So in the free edition, the audio genuinely has nowhere to go. (Their pricing FAQ says cloud and custom transcription providers are a Pro thing.)
• Summarises with a built-in local model, with Ollama, or with your own key for Claude, Groq, OpenRouter, OpenAI or any OpenAI-compatible endpoint. Only the transcript text goes out on the key route.
• Imports audio and video files and re-transcribes old meetings with a different model or language. Both free.
• Keeps the recording by default. auto_save is true in their recording preferences.
• Analytics through PostHog, off unless you turn it on. That became the default in v0.4.0. It used to be the other way round, and they changed it, which is the right direction to change it in.

That's a real product. If what you want is "a transcript and a summary of my call, without anyone's cloud in the middle," Community does it, for free, on two operating systems.

#The decision in the source that the feature lists miss

Okay. This is the part my Bias has been waiting for, though I'll argue it's not really about bias, because it's the thing I'd want to know if I were choosing.

Meetily's audio pipeline lives in pipeline.rs. It takes your microphone and the system audio, lines them up in 50-millisecond windows, and mixes them into one signal. That single mixed signal is what goes to voice detection and transcription, and it's also what gets saved as the recording. The comment in the code is pleasingly frank about it: only the mixed audio is saved, "mic + system properly mixed."

That's a completely reasonable engineering choice. One stream is simpler, it's cheaper to transcribe, and their mixer does careful things to stop the two sources clipping each other.

But think about what it costs. Once your voice and the call are one waveform, the transcript has no idea which side of the call a sentence came from. There's no "You:" and no "Them:". It's one long wall of words, and on a two-person sales call that's the difference between "the client agreed to the price" and "I said the price and nobody objected." Which, if you've ever had to reread a transcript after a client remembered a deadline differently, are not the same sentence AT ALL.

Humla made the opposite call. Mic and system stay two separate streams from capture to transcript, never mixed down, so "you vs them" is known structurally before any AI gets involved. Then on-device speaker diarization splits "them" into individual people. Free, in the MIT app.

Meetily's answer to who-said-what is Pro, and Pro's version looks good on paper: speaker identification that runs on-device, live while you record and on imported audio, with renaming and merging duplicate speakers. I haven't run it, because it's behind a licence key, so I'm going on their description. But that's where it lives.

(Small aside for anyone else who reads GitHub descriptions: Meetily's repo description currently says "on-device Whisper/Parakeet transcription and speaker diarization." The README and the pricing table both say speaker identification is Pro. The pricing table is the one to trust. Our own best local notetakers post said Meetily hadn't shipped diarization at all, which was true in July and isn't now, and I've fixed it.)

#What Pro costs and what's in it

From meetily.ai/pricing and meetily.ai/pro, 10 November 2026:

The $10 is described as 60% off a $25/month regular price. There's a 14-day trial with no card, and a 30-day money-back guarantee, which is generous. Current Pro version is 1.11.2.

The line in that table I'd want a buyer to notice is the licence row. Pro is not open source. Meetily is open-core: the MIT edition is real and free forever (their words, and I believe them), but the features that make it a team tool live in a closed product on top. That's a legitimate way to fund an open-source project, and honestly it's probably a smarter business than mine. It does mean "Meetily is open source" and "the Meetily you'd actually pay for is open source" are different sentences.

#Where Meetily wins, plainly

My Bias is going to sit this section out. He can wait in the car.

Windows. Not a small thing. Humla is Mac-only, Apple Silicon, and will be for the foreseeable future. If anyone on your team is on Windows, Meetily is the local, open-source option, and I'm not going to bury that under a "but."

Parakeet. It's fast and accurate, and Meetily ships it as the recommended engine. Humla uses Whisper on-device (plus a Norwegian-tuned model for Norwegian meetings, because that's what we speak at work). I don't ship Parakeet.

Community and momentum. A hundred times the stars, a steady flow of new contributors, a Discord, a subreddit. If the bus factor of a small project worries you, and for a tool that holds your meeting history it should, Meetily has the much better answer.

Import and re-transcribe, free. Humla imports recordings too, but Meetily's "enhance" flow for re-running a meeting with a different model is a nice touch.

Transcription that can't leave. Humla offers OpenAI, Deepgram and Groq transcription alongside on-device Whisper, and I think that's useful. But a tool where the cloud option literally isn't in the free build is a simpler promise to make to a nervous IT department.

#Where Humla wins, with the Bias back in the room

He's back. He brought a list. I've cut it down to the things I think actually matter:

• Who said what, free. Separate streams plus on-device diarization, in the MIT app. Meetily has this only in paid, closed Pro.
• Chat over your notes, shipped. Ask questions across your meeting history with answers that cite the meetings they came from. Meetily lists chat as "Soon" on every tier.
• MCP and webhooks, free and open source. Humla's MCP server lets Claude Code or Codex read your notes, and Send to webhook… pushes a note to your own automation. Meetily has both, in Pro.
• Audio gone by default. Humla deletes the audio once the transcript exists unless you turn retention on. Meetily keeps the recording by default. Neither is wrong, but if a recording on disk is the thing your legal team worries about, defaults matter.
• One licence for everything. The whole Humla app is MIT. The only paid thing is optional team sync through Humla Cloud, $5/user/month, which is a service rather than a locked feature.

And the honest weaknesses, so the list isn't a sales sheet: Mac only, Apple Silicon only, no mobile app, no CRM integrations, and a community you could fit in a meeting room. If I get hit by a bus, Humla has a problem Meetily doesn't.

#So, Meetily or Humla

Pick Meetily if anyone in the picture uses Windows, if you want the biggest and most active open-source project in this space, or if you want a free tool where transcription physically can't go to the cloud and you don't need to know who said what.

Pick Humla if you're on a Mac and want speaker labels, chat and the MCP server without a licence key, or you want the whole thing to be one MIT codebase you can read.

The full grid is on the Humla vs Meetily page, and if you're weighing the wider open-source field, this roundup has five more.

#Verdict

Meetily is the review I expected to find hardest, and it was, just not for the reason I thought. I assumed the hard part would be resisting the urge to dunk on a competitor. It turned out the hard part was that there's nothing to dunk on. It's a well-run project making a privacy promise it actually keeps, with a business model that's more sustainable than mine, on more platforms than mine.

The real difference between us isn't local versus cloud, or open versus closed. It's one engineering decision about whether two people's voices become one waveform, and everything downstream follows from it: where the speaker labels come from, what's free and what's Pro, what a transcript can prove later.

My Bias, for the record, thinks I've been far too nice. He's probably right. He usually is about the star count.

---

Meetily facts from github.com/Zackriya-Solutions/meetily (README, release notes for v0.4.0 and v0.4.1, and the source at commit 07c4dfd: frontend/src-tauri/src/audio/pipeline.rs, recording_preferences.rs, src/components/TranscriptSettings.tsx, PRIVACY_POLICY.md), the GitHub API, and meetily.ai/pricing and /pro, all on 10 November 2026. Meetily Pro was not run; its features are as Meetily describes them. Other reviews: Granola, Otter.ai.