#1★ TOP PICK
Meetily
MIT-licensed meeting notes that record, transcribe and summarise entirely on your own machine.
95
OPEN SOURCEMITSELF-HOSTLOCAL-FIRST
Meetily is a desktop meeting assistant from Zackriya Solutions that captures your device audio, transcribes it in real time with Whisper or NVIDIA Parakeet, and writes summaries through a model of your choosing — Ollama on the same computer if you want the whole loop local, or Claude, Groq, OpenRouter or any OpenAI-compatible endpoint if you don't. Recordings, transcripts and models stay on the machine. The Community Edition is MIT-licensed and free; a separate Pro licence adds diarization, templates and richer exports, so it is honest to call this open-core rather than purely open-source.
Strengths
- +Audio, transcripts and summaries can stay entirely on your device — the project's own documentation states all processing is local
- +Packaged installers for macOS and Windows, with Metal and CoreML enabled on Mac and a Vulkan Whisper build on Windows; the README also lists Linux
- +Your choice of model at both ends: Whisper or Parakeet for transcription, Ollama locally or Claude/Groq/OpenRouter for summaries
- +Actively developed and widely used — v0.4.1 released 2026-09-12, roughly 31,000 GitHub stars
- +The free Community Edition is genuinely usable, not a demo: live transcription, imports and local summaries are all in it
Trade-offs
- −No bot joins the call, so it records the audio on the machine you are meeting from — it cannot cover a meeting you attend from your phone
- −Speaker diarization, custom summary templates and PDF/DOCX/Markdown export sit in the paid Pro licence, not the free edition
- −The Windows installer needs an AVX2-capable x64 CPU, and CUDA acceleration requires building from source with an NVIDIA toolchain
- −You own the maintenance: model downloads, updates and disk space are yours, and there is no mobile app or web workspace
- −Transcription speed and accuracy depend on your hardware and the model size you pick — test it on a real recording before committing
Community Edition free and open source — local transcription, local AI summaries, audio/video import, retranscription. Optional Pro licence is $120 per user per year, shown as $10/user/month billed annually against a $25 list price, with a 14-day trial; Pro adds speaker diarization, a custom AI model connector, custom summary templates, PDF/DOCX/Markdown exports and MCP/webhook/CLI integrations. Enterprise volume licensing is quoted. Verified on meetily.ai/pricing 2026-09-25. #2
Vibe
Fully offline desktop transcription for recordings, with an unusually long export list.
98
OPEN SOURCEMITSELF-HOSTLOCAL-FIRST
Vibe is a small MIT-licensed desktop app that transcribes audio and video files on your own machine — no account, no cloud, no per-minute billing. It runs on Windows, macOS and Linux with GPU acceleration on Nvidia, AMD and Intel hardware, batches whole folders, and writes out SRT, VTT, TXT, HTML, PDF, JSON or DOCX. It is the right tool if what you actually need is transcripts of interviews, lectures or recorded calls rather than a live meeting assistant.
Strengths
- +Completely offline: nothing is uploaded, which removes the data question rather than answering it
- +Windows, macOS and Linux, with Vulkan or CoreML acceleration across Nvidia, AMD and Intel GPUs
- +Batch transcription of multiple files, and it can pull audio from sites such as YouTube and Vimeo
- +Exports SRT, VTT, TXT, HTML, PDF, JSON and DOCX, so a transcript is never trapped in the app
- +Maintained and current — v3.2.2 released 2026-09-05
Trade-offs
- −Not a meeting assistant: it works on recordings, so there is no calendar, no live call capture and no action-item output
- −The feature list does not include AI summaries or speaker diarization, which is a real part of what Otter produces
- −Effectively a single-maintainer project — healthy today, but that is the risk you are accepting
- −You still have to get the audio: capturing a Zoom or Teams call is a separate step you arrange yourself
- −Accuracy and speed follow your hardware and the model you pick, as with every local Whisper setup
Free and open source. No accounts, no paid tier and no per-minute charge — you provide the computer. The project accepts donations on its site. #3
Anarlog
Local-first AI meeting notepad that takes notes without a bot joining the call.
94
OPEN SOURCEMITSELF-HOSTLOCAL-FIRST
Anarlog is an MIT-licensed desktop notepad, built by Fastrepl, that listens to your device audio rather than joining your meeting as a participant. Sessions, notes and transcripts live in a local SQLite database with recordings as ordinary files on disk, and you choose where the intelligence comes from: an on-device model, your own API key, or an OpenAI-compatible local server such as Ollama or LM Studio. Hosted transcription, encrypted sync and sharing exist as a paid tier, but nothing depends on them. It was called Hyprnote until the project renamed itself in 2026, which is worth knowing when you search for guides.
Strengths
- +Ships macOS (Apple Silicon and Intel), Windows and Linux builds in the same release — desktop_v1.4.27 was published 2026-09-25
- +Nothing joins your call and nothing appears in the participant list; it also stays hidden while you share your screen
- +Your data is readable without the app: local SQLite plus plain recording files, with Markdown export
- +Bring your own AI — hosted provider, your own key, or a local OpenAI-compatible server; cloud sync is off unless you turn it on
- +Very active development, with releases landing almost daily
Trade-offs
- −The rename from Hyprnote means older bookmarks, tutorials and blog posts point at a name and repository that now redirect
- −Its README states the team is now building a separate product, char, while the community app stays MIT-licensed and maintained — worth watching if you are betting a workflow on it
- −Enterprise components are source-visible and commercially licensed, not MIT — the free, MIT part is the community desktop app
- −Built-in on-device transcription models are described as available on supported Macs, so Windows and Linux users may need their own keys or a local server
- −Like every device-audio notetaker, it only covers meetings you attend on that computer
Free tier: unlimited on-device transcription, local recordings, bring your own speech and LLM keys, notes/folders/templates, exports, and a local API with CLI, MCP and webhooks. Pro is $15/month or $150/year and adds hosted transcription and LLM, system-wide dictation, better speaker identification, end-to-end encrypted sync across 3 devices and note sharing. Team is $20/person/month or $200/person/year; Enterprise is custom with SSO and SCIM. Verified on anarlog.so/pricing 2026-09-25. #4
Scriberr
Self-hosted transcription server with an API, for a household or a small team.
92
OPEN SOURCEMITSELF-HOSTLOCAL-FIRST
Scriberr is an MIT-licensed, offline transcription application built for self-hosters: you run it on your own server, upload audio, and it transcribes with NVIDIA Parakeet, Canary or Whisper models and word-level timing. Because it is a server rather than a desktop app, several people and devices can share one install, and it exposes an API for automation. It is the right shape if transcription is a service in your infrastructure rather than something on one laptop — with a maintenance caveat worth reading before you commit.
Strengths
- +Runs entirely on infrastructure you control, with no third-party cloud in the path
- +Modern model choice — NVIDIA Parakeet and Canary as well as Whisper — with word-level timing
- +One install serves several people and devices, which a desktop app cannot do
- +Has an API, so transcription can be wired into other tools you run
Trade-offs
- −Development was openly paused for much of 2026 while the maintainer changed jobs and moved; the last tagged release is v1.2.0 from December 2025, though commits have resumed (repository last pushed 2026-09-20)
- −The documentation links published on its own site (scriberr.app/docs) returned 404 when we checked on 2026-09-25, so setup currently means reading the repository
- −A single-maintainer project, as the maintainer states plainly himself
- −It transcribes uploaded audio rather than sitting in your meetings — no live meeting capture and no action items
- −Self-hosting is real work: a machine, storage, updates, and enough CPU or GPU to be worth using
Free and open source; you supply the hardware. The project takes donations through Ko-fi and has no paid tier. #5
Granola
The cheaper polished cloud notetaker — the honest answer when the objection is price, not privacy.
18
SOURCE-AVAILABLEProprietary (hosted service)
If what prompted the comparison is the invoice rather than where the audio goes, Granola is worth knowing about: a hosted AI notepad that captures device audio without adding a bot to the call, at $14 per user per month against Otter Pro's $16.99 monthly rate. It is proprietary and cloud-based — no self-hosting, no local models — and we list it because for plenty of readers that is the right trade, stated plainly. Check the arithmetic before switching for cost alone: Otter billed annually is $8.49 per user per month, which is cheaper than Granola.
Strengths
- +$14/user/month undercuts Otter Pro's $16.99 monthly rate, with a free tier to try first
- +No bot joins the meeting — it captures device audio, so nothing appears in the participant list
- +Desktop, iOS, Android and Apple Watch, with nothing for you to run or maintain
- +Its plan page offers opting out of model training on the free tier, and org-wide on Enterprise
Trade-offs
- −Fully hosted and proprietary: no self-hosting, no local models, no offline mode
- −Your meeting record lives in Granola's cloud, so the question that sends people looking for an alternative to Otter applies here too
- −API access is listed from the Business tier upward
- −Otter's annual rate of $8.49/user/month is lower — the saving only exists against Otter's monthly billing
- −The Sovereignty Score of 18 reflects a closed hosted service; it is a transparency lens, not a verdict on how good the product is
Basic free with limited meeting history; Business $14/user/month; Enterprise $35/user/month. Verified on granola.ai/pricing 2026-09-25.