Meetily is a free, open-source AI meeting transcription tool that runs entirely on your local machine. No bot joins your call. No audio leaves your device. You transcribe meetings with Whisper or Parakeet on your own hardware, summarize with local models or your own API keys, and store everything locally. The Community Edition is MIT-licensed and costs nothing. A Pro tier at $10/user/month adds speaker identification, GPU-accelerated transcription on Windows, custom summary templates, and advanced exports. With over 25,900 GitHub stars and 369,000-plus downloads as of July 2026, it has become the most-watched open-source project in the local-meeting-tools space.
What makes it interesting is the architecture. It is built with Tauri (Rust shell) and Next.js, making it a lightweight desktop app instead of a browser-based service. For transcription, it bundles Whisper medium and Whisper large-v3 for users with sufficient hardware, or Parakeet for faster, lighter processing. For summaries, you choose between local models via Ollama (Gemma 3 4B/1B, Qwen) or external APIs through a BYOK (Bring Your Own Key) system that supports OpenAI, Anthropic, Gemini, Groq, Mistral, OpenRouter, and Azure OpenAI. The audio recording, transcript text, and summary all stay on your local filesystem. The BYOK path sends transcript text only — never audio.
The reviewer testing it on YouTube gave it a 4-out-of-5 for the local AI notetaker use case. The two knocks: no speaker diarization in the transcript view (only in the Pro summary), and a slight processing delay on lower-end hardware. Those are real limitations, not deal-breakers, and they tell you exactly who this tool is for: people who need accurate, private meeting notes without a cloud dependency.
How Meetily Works Under the Hood
Meetily captures system audio during meetings — the same audio you hear through your speakers — without joining the call as a visible bot. It records locally, transcribes locally using the model you select, and stores everything as files on your disk. The architecture has three layers:
- Audio capture: Uses your operating system's audio loopback to record what comes through your speakers and microphone. No virtual audio cable required. Works with Zoom, Microsoft Teams, Google Meet, Webex, FaceTime, Slack calls, Discord — any app that plays audio through your system.
- Transcription (local): Runs Whisper or Parakeet models on your device. Whisper medium works on machines with 8GB RAM. Whisper large-v3 (the most accurate model, at 2.1% word error rate on clean audio per benchmarks) requires roughly 10GB VRAM or 16GB system RAM. Parakeet is faster and lighter, suitable for real-time transcription without a discrete GPU.
- Summarization (local or BYOK): You pick the language model that generates your meeting notes. Run Gemma 3 4B or Qwen locally through Ollama for a fully offline setup. Or bring your own API key from OpenAI, Anthropic, Google, Groq, Mistral, OpenRouter, or Azure. When you use BYOK, only the transcript text is sent — the audio file never leaves your machine.
That last point is the one Meetily's own pricing page emphasizes repeatedly. "Audio and recordings never leave your device." If you use a local model for summaries, nothing leaves your device at all. This is the architecture that makes it suitable for healthcare, legal, financial, and government use cases where cloud-based tools cannot pass compliance review.
Whisper Benchmark: How Accurate Is Local Transcription?
The accuracy question is the one most people ask first. Can open-source Whisper match cloud transcription services like Otter or Fireflies?
The short answer: on clear audio with a decent microphone, the gap is negligible. On messy audio with heavy accents or cross-talk, cloud services still have an edge — but it is closing.
Whisper large-v3, the model Meetily can optionally run, achieves a 2.1% word error rate on clean audio and 3.8% on real-world audio, according to published benchmarks. That puts it on par with Google Speech-to-Text (2.3% clean WER) and ahead of both Deepgram (2.5%) and AssemblyAI (2.6%). For meeting transcription specifically — the use case Meetily targets — large-v3 hits 4.2% WER, which handles overlapping speech well.
The trade-off is compute. Whisper large-v3 needs roughly 10GB of VRAM for GPU-accelerated processing or 16GB of system RAM for CPU mode. On a GPU like an RTX 4090, it processes audio at a 0.15x real-time factor — meaning a 60-minute meeting transcribes in under 10 seconds. On CPU alone, that same meeting could take 8.5 hours. This is why Meetily offers multiple model sizes: Whisper medium (5GB VRAM, 3.8% clean WER) for people without high-end hardware, and Parakeet for real-time transcription on modest machines.
For most one-on-one calls and small group meetings with reasonable audio quality, the difference between local Whisper and cloud transcription is not something you will notice. For recorded meetings with background noise or many overlapping speakers, the cloud services may still pull ahead.
Meetily Community vs. Pro: What You Get for Free
The Community Edition is the free tier, and it is not a crippled trial. It includes:
- Real-time transcription with Whisper or Parakeet on your device
- AI-powered summaries using local models (via Ollama) or your own API keys
- Audio and video import in 10 formats for re-transcribing old recordings
- Local storage of all recordings, transcripts, and summaries
- Basic sharing via copy
- Community support through Discord and GitHub
What Pro ($10/user/month, billed annually at $120) adds on top:
- Speaker identification (who said what) — the biggest gap in the free tier
- Enhanced accuracy models — larger Whisper and Parakeet variants with quality presets for noisy meetings
- Windows GPU support — GPU-accelerated transcription on Windows for faster processing of large models
- Custom AI model connector — connect to any OpenAI-compatible endpoint
- Custom summary templates — build your own note template and reuse it across meetings (Community gets limited templates; Pro gets 6 built-in plus custom)
- Auto-detect meetings — automatically start recording when a meeting begins, without manual trigger
- Advanced exports — PDF, DOCX, and Markdown export for sharing meeting notes with your team
- Priority email support
The 14-day Pro trial requires no credit card. After the trial, your data stays intact and you fall back to Community features. The license is per-device — one Pro seat covers one machine.
Meetily Pricing Table
| Feature | Community (Free) | Pro ($10/user/mo, annual) | Enterprise (Custom) |
|---|---|---|---|
| Local transcription | Yes | Yes | Yes |
| AI summaries | Local + BYOK | Local + BYOK | Local + BYOK |
| Speaker identification | No | Yes | Yes |
| Enhanced accuracy models | No | Yes | Yes |
| Windows GPU support | No | Yes | Yes |
| Custom summary templates | Limited | 6 built-in + custom | Org-wide library |
| Advanced exports (PDF, DOCX, MD) | No | Yes | Yes |
| Auto-detect meetings | No | Yes | Yes |
| Priority support | Community (Discord, GitHub) | Email priority | Dedicated + SLA |
| Price | Free, MIT license | $120/year | Custom |
How Meetily Compares to Cloud Meeting Notetakers
The meeting notetaker market splits into two camps. Cloud tools (Otter.ai, Fireflies.ai, Fathom) send your audio to their servers for processing. Local tools (Meetily, Anarlog, Shadow, Talat) process on your device. The trade-off is convenience versus control.
Cloud tools are easier — install once, it works, your notes sync across devices. But they require sending your meeting audio to a third party, they show a visible bot in your call, and their pricing scales per seat. Local tools need more setup and local hardware, but your audio never leaves your machine.
Here is how Meetily stacks up against the most popular options.
Meetily vs. Otter.ai
Otter.ai is the cloud incumbent. Its Pro plan costs $16.99/month per user and its Business plan $30/user/month. Otter sends a visible bot into your meeting, transcribes in the cloud, and stores your notes on Otter's servers. It has strong integrations with Salesforce, HubSpot, and Slack.
Meetily's Community Edition is free and processes everything locally. Otter is easier to set up and has broader integrations. Meetily is better for anyone who cannot or will not send meeting audio to a third party — legal, healthcare, financial, government, or anyone under an NDA.
Meetily vs. Fireflies.ai
Fireflies.ai offers a free tier with 800 minutes of storage, then $10/month for Pro and $19/user/month for Business. Like Otter, it sends a bot into your meeting and processes audio in the cloud. Fireflies has broader CRM integrations and conversation-intelligence features (sentiment analysis, talk-time tracking) that Meetily does not match.
Meetily's Pro plan is the same price as Fireflies Pro ($10/month), but Meetily's advantage is privacy and the free Community tier. Fireflies is the better pick if you need CRM integration and conversation analytics. Meetily wins if privacy is non-negotiable.
Meetily vs. Granola
Granola is a Mac-only, bot-free meeting notetaker — like Meetily, it captures system audio without joining your call. But there the similarity ends. Granola sends your audio to Deepgram for transcription, processes your transcript through GPT-4o or Claude for summaries, and stores notes on AWS servers. There is no offline mode and no option to run local models. Granola is free for a basic tier, then $14/user/month for Business and $35/user/month for Enterprise.
Meetily is the open-source alternative that does what Granola does not: keeps everything on your device. If you like Granola's bot-free UX but cannot accept cloud processing, Meetily is the closest substitute — and it runs on Windows too, while Granola is Mac-only.
Meetily vs. Fathom
Fathom is free for individuals, with a Team plan at $19/month. It is a cloud-based bot tool — a visible bot joins your meeting and processes audio server-side. Fathom is arguably the strongest free cloud option: unlimited transcription on the free tier, good integrations, and polished summaries.
Meetily is not trying to compete with Fathom on integrations or the breadth of features. It competes on the architectural guarantee: your audio does not leave your machine. For a consultant recording client calls under NDA, that matters more than a Slack integration.
Pricing Comparison Summary
| Tool | Free Tier | Starting Price | Bot-Free | Local Processing |
|---|---|---|---|---|
| Meetily Community | Yes (MIT license) | $10/user/mo (Pro) | Yes | Yes |
| Otter.ai | 300 min/mo | $16.99/mo | No | No |
| Fireflies.ai | 800 min storage | $10/mo | No | No |
| Granola | Free (limited) | $14/user/mo | Yes | No (cloud transcription) |
| Fathom | Unlimited | $19/mo | No | No |
System Requirements and Setup
Meetily runs on Windows 10+ and macOS 11+. The Community Edition can also be built from source on Linux, though Pro does not yet ship native Linux installers.
Minimum specs:
- 8GB RAM
- Any modern CPU (Intel, AMD, or Apple Silicon)
Recommended specs for Whisper large-v3:
- 16GB RAM + discrete GPU with 10GB VRAM, or
- Apple Silicon Mac (M1 or later with 16GB unified memory)
For real-time transcription without a discrete GPU:
- Use Parakeet or Whisper medium — both run on 8GB RAM machines
- Whisper large-v3 on CPU will work but will be slow (not real-time)
The setup flow: download the app from meetily.ai, open it, select your transcription model (Parakeet for speed, Whisper medium for balance, Whisper large-v3 for max accuracy if your hardware supports it), choose your summary model (local via Ollama or BYOK), and start a meeting. Meetily captures audio automatically — no one in the call sees a bot.
Where Meetily Falls Short
The 4-out-of-5 reviewer rating tells the story. Meetily does most things well, but it has real gaps:
No speaker diarization in the Community Edition. The transcript appears as one block of text. Pro adds speaker identification in the summary, but if you need per-speaker attribution in the raw transcript, the free tier does not provide it. Competitors like Otter and Fireflies handle this on their free tiers (because they process in the cloud).
No cloud sync or cross-device access. Your transcripts live on the machine where the meeting was recorded. That is by design — it is the privacy guarantee — but if you work across a desktop and a laptop, you need to export notes manually or use the file system to sync them.
No mobile app. Meetily is a desktop application. There is no iOS or Android companion. If you need meeting notes on your phone, you export from the desktop app.
Linux is Community-only. Pro does not ship native Linux installers. A native Linux build is on the roadmap, and Community can be built from source today, but Pro users on Linux are out of luck for now.
Calendar integration is "coming soon." Auto-detect meetings works, but true calendar integration (reading your schedule, pre-joining meetings) is listed as "Soon" on the pricing page, not "Available." Competitors have this today.
Processing delay on lower-end hardware. On a machine without a GPU, Whisper large-v3 takes significantly longer than real-time to transcribe. You will wait. Parakeet or Whisper medium avoid this, at the cost of some accuracy.
Who Should Use Meetily
Best fit: Privacy-sensitive professionals — lawyers, doctors, financial advisors, government workers, consultants under NDA, developers working on confidential projects. Anyone who cannot or will not send meeting audio to a third-party cloud service.
Also good for: Developers and technical users who want an open-source tool they can audit, modify, or self-host. Teams that want local meeting notes without per-seat subscription costs (Community Edition is free, MIT-licensed).
Not the right fit: Teams that need deep CRM integrations (Salesforce, HubSpot pipeline syncing) — cloud tools do this better today. Anyone who needs meeting notes synced across multiple devices — Meetily is desktop-only with no cloud sync. Users who need speaker diarization on a free plan — Otter and Fathom offer this on their free tiers, Meetily free does not.
The Verdict
Meetily earns its 25,900 GitHub stars by solving a problem the cloud tools refuse to address: keeping your meeting audio under your control. The Community Edition delivers real value — transcription, summaries, local processing — for $0. The Pro tier at $10/month is priced to compete with Fireflies and Otter while offering a privacy architecture they cannot match.
The tool is not perfect. The missing speaker diarization on the free tier, the lack of cloud sync, and the absence of a mobile app are all real limitations. But for the specific use case — private, local, free meeting transcription — Meetily is the best open-source option available in 2026. If you have been paying $17/month to Otter or $14/month to Granola and your meetings are sensitive enough that you would rather they never touch a cloud server, download the Community Edition and see if the accuracy is good enough for your setup. If it is, you just cut a recurring subscription from your budget.
FAQ
Is Meetily really free?
Yes. Meetily Community Edition is free and open source under the MIT license. It includes real-time transcription, AI-powered summaries, local processing, and audio/video import. There is no trial period, no credit card, and no time limit. The Pro tier adds features like speaker identification and advanced exports for $10/user/month billed annually.
Does Meetily send audio to the cloud?
No. Transcription always runs locally on your device. Audio and recordings never leave your machine. For summaries, you choose: run a local model through Ollama (nothing leaves your device) or use Bring Your Own Key (BYOK) to send only the transcript text to an AI provider you choose (OpenAI, Anthropic, Gemini, etc.). Audio is never uploaded.
Does Meetily work with Zoom, Teams, and Google Meet?
Yes. Meetily captures system audio, so it works with any meeting platform — Zoom, Microsoft Teams, Google Meet, Webex, FaceTime, Slack calls, Discord, and any app that plays audio through your speakers. No bot joins your meeting; other participants see nothing.
What hardware do I need for Meetily?
Minimum: 8GB RAM on Windows 10+ or macOS 11+. For Whisper large-v3 (the most accurate transcription model), you need roughly 16GB RAM plus a discrete GPU with 10GB VRAM or an Apple Silicon Mac (M1 or later). Whisper medium and Parakeet models run on 8GB RAM machines. GPU is strongly recommended for real-time transcription with large models.
What is the difference between Meetily Community and Pro?
Community (free, MIT license): local transcription, AI summaries (local or BYOK), audio import, basic sharing. Pro ($10/user/month annual): adds speaker identification, enhanced accuracy models, Windows GPU support, custom summary templates, auto-detect meetings, advanced exports (PDF/DOCX/Markdown), and priority email support. A 14-day free trial lets you test Pro features with no credit card.
Can I use Meetily offline?
Yes. If you use local transcription (Whisper or Parakeet) and local summarization (via Ollama), Meetily works fully offline. You need internet only to download the app and the AI models. After that, no connection is required. BYOK summarization does require internet to reach your API provider, but audio still never leaves your device.
How does Meetily compare to Granola?
Both are bot-free (no visible bot joins your call). Granola sends audio to Deepgram for transcription and processes summaries through GPT-4o or Claude — it is a cloud tool with local audio capture. Meetily processes everything locally. Granola is Mac-only; Meetily runs on Windows and macOS (Community also on Linux). Granola costs $14/user/month for Business; Meetily Community is free and Pro is $10/user/month.
Is Meetily good for teams?
Community Edition is suitable for individual use. Pro adds per-user licensing at $10/user/month and includes features useful for teams: auto-detect meetings, advanced exports, and priority support. Enterprise pricing offers volume licensing, dedicated support, custom deployment, and SSO integration (SSO is on the roadmap). There is no mobile app and no cloud sync, so team collaboration relies on exported files shared through your existing tools.
Discussion
0 comments