On-device: audio never leaves your machine. No bot, no cloud.
Speechmark is the newest local-first meeting-notes app, and it earns the label. A menu-bar Mac app, no account, no login, audio never leaves your Mac. It transcribes on Apple Foundation Models by default, with an optional cloud model that's text-only and bring-your-own-key. There's a Claude Desktop MCP connector for piping transcripts into a local AI workflow. Speaker ID uses a confirm-name flow. Action items are supported. Local LLM summaries run via Ollama or Apple Foundation. Bring-your-own keys cover OpenAI and Anthropic. Usage stats are opt-in and off by default. It's bootstrapped by ChakraBit Solutions, and the privacy positioning is the cleanest in the local-first field.
The pricing gets noticed. Speechmark is free during its public beta, no credit card, and the launch price is a one-time $49 to $79 per Mac with no subscription. For a solo user who already owns a Mac and doesn't want a recurring line item, that's a real shape, and most of the cloud incumbents don't offer it.
What Speechmark does well is the menu-bar capture and the MCP connector. The menu-bar app means capture is one click from anywhere on macOS. The Claude Desktop connector means transcripts flow into a local AI assistant without a cloud detour. For a Mac-only workflow that already runs through Claude Desktop, that's a tight loop, and the one-time price makes it cheap to try.
The first wall is the platform. Speechmark is macOS only, and specifically Apple silicon. There's no Windows build, no Linux build, and no stated plan for either. If your team runs mixed OS, or you're the one Linux user in a Mac shop, Speechmark isn't an option, and the menu-bar polish doesn't fix that.
The second wall is timing. Speechmark has no live transcript. It records, then processes after you stop. The transcript and the summary arrive together, post-call. For a meeting where you want to glance at the transcript mid-call, catch a missed name, or pull up an action item as it's spoken, that's a structural limit. You get a clean post-hoc record, not a live one.
The third wall is proof. Speechmark keeps audio local and says so, but it doesn't ship a privacy network monitor. There's no in-app view that shows zero outbound connections during a meeting. For a buyer who wants to demonstrate the local-first claim to a compliance reviewer, a client, or their own certainty, the absence of a monitor is a gap. The claim is true. The proof isn't in the product.
The integration surface is also small. The named integration is Claude Desktop via MCP. There's no CRM sync, no mobile app, no shared cloud workspace. For a solo Mac user that's fine. For a team that wants notes pushed into a pipeline, it's a thinner surface than the cloud incumbents offer.
People who switch off Speechmark tend to fall into one of three groups. Some need Windows or Linux support and can't use a Mac-only tool. Some want a live transcript they can read while the meeting happens. Some need an in-app network monitor to prove zero outbound connections, not just take the claim on trust. The list below speaks to each group, which is why it isn't ranked on a single axis.
Ranked by fit for local, bot-free meeting notes. Clearminutes is the local-first anchor at the end.
Otter.ai puts a bot in every call and sends audio to its servers. Clearminutes runs on your machine.
Best for: sales teams that want CRM-synced cloud notes and a polished shared workspace. A bot joins every call and audio goes to Otter's servers. The free tier caps at 300 minutes a month, with no on-device or offline option.
Price: $0/forever
Otter.ai is best for sales teams who want CRM-synced cloud notes and a polished shared workspace. A bot joins every call and audio goes to Otter's servers, with a free tier capped at 300 minutes a month and no on-device or offline option. Strong on Zoom, Teams, Salesforce, and HubSpot integration, weak on privacy and native cross-platform apps. Pick it when team collaboration matters more than keeping audio on the device.
Meeting transcription without the awkward "Fireflies has joined" announcement or monthly subscription.
Best for: teams that want a rich searchable transcript archive and deep CRM integrations with Salesforce and HubSpot. A cloud bot joins every call and audio is stored on Fireflies.ai's servers. The free tier is limited to 800 minutes of storage per seat.
Price: $0/forever
Fireflies.ai is best for teams who want a rich searchable transcript archive and deep CRM integrations like Salesforce and HubSpot. A cloud bot joins every call and audio is stored on Fireflies.ai's servers, with the free tier limited to 800 minutes of storage per seat. Strong on search and CRM workflows, weak on privacy and offline use. Pick it when searchable team transcripts matter more than local processing.
Fathom is a free Zoom-only cloud notetaker. Clearminutes works with any audio source, runs locally on Mac and Windows, and doesn't need Zoom.
Best for: teams on Zoom who want a free, simple cloud notetaker with unlimited recordings and 28-language support. A bot joins every call, coverage is Zoom-only, and audio is uploaded to Fathom's servers. There's no on-device transcription path and no offline option.
Price: $0/forever
Fathom is best for teams on Zoom who want a free, simple cloud notetaker with unlimited recordings and broad 28-language support. A bot joins every call, coverage is Zoom-only, and audio is uploaded to Fathom's servers with no on-device transcription path. Strong on free-tier generosity and language coverage, weak on platform breadth and privacy. Pick it when you live on Zoom and want free cloud notes.
Jamie is bot-free but summarises in the cloud on Mac and Windows. Clearminutes keeps the summary on-device too, with a real free tier and Linux support.
Best for: individuals who want a free, bot-free local summary on Mac and Windows with export to Notion, Slack, and HubSpot. The free tier is limited and the paid plan is £25/month for fuller features. The feature set is lighter than the cloud incumbents.
Price: £0/forever
Jamie is best for individuals who want a free, bot-free local summary on Mac and Windows with Notion, Slack, and HubSpot export. Processing is partial — offline is available but summaries are not fully on-device. The free tier is limited and the paid plan is £25 a month. Strong on bot-free capture and export, weak on Linux, a real free tier, and a local LLM. Pick it for solo Mac or Windows use.
Granola is a polished Mac-only drafting app that processes audio in the cloud. Clearminutes runs on-device, on Mac and Windows, with a one-off Pro price.
Best for: Mac users who want a polished meeting summary app that drafts notes and exports to Notion, Slack, and HubSpot. It's Mac-only with cloud-only audio processing, a limited free tier, and a £14–35/month subscription. The integration surface is smaller than the cloud incumbents.
Price: £0/forever
Granola is best for Mac-only users who want a polished cloud-native drafting workflow with custom templates, chat-with-meetings, action items, cross-meeting search, and speaker diarization, plus Notion, Slack, and HubSpot export. Audio is processed on Granola's servers with no local option, and the paid plans run £14–35 a month. Strong on the Mac drafting experience, weak on Windows, Linux, offline, and a real free tier. Pick it for Mac-only work in those three tools.
Krisp pioneered noise cancellation. Clearminutes is a complete AI meeting assistant: transcription, summaries, action items, and task sync.
Best for: teams that need noise cancellation first and a meeting assistant second. Transcription is a cloud add-on, not the core product. There's a 14-day trial then a limited free tier, 16-language support, and an $8–15/month paid plan for full features.
Price: $0/forever
Krisp is best for teams who need noise cancellation first and a meeting assistant second. Transcription is a cloud add-on rather than the core product; noise cancellation runs locally but transcription does not. A 14-day trial gives way to a limited free tier, with a paid plan at $8–15 a month for full features across 16 languages. Strong on noise removal, weak on on-device transcription and offline notes. Pick it when call audio is noisy.
tl;dv records Google Meet, Zoom, and Teams in the cloud with a bot in the call. Clearminutes keeps the recording on your machine and works offline.
Best for: teams on Google Meet, Zoom, and Teams who want a free-tier cloud recorder with 30+ language transcripts and Chrome-based capture. A bot joins every call and audio is uploaded to tl-dv's servers. The paid plan is $20–25/month, with no on-device transcription option.
Price: $0/forever
tl;dv is best for teams on Google Meet, Zoom, and Teams who want a free-tier cloud recorder with 30+ language transcripts and Chrome-based capture. A bot joins every call and audio is uploaded to tl;dv's servers, with a $20–25 a month paid plan and no on-device transcription option. Strong on multi-platform recording and language coverage, weak on privacy and offline use. Pick it for free cloud recording across meeting platforms.
Avoma is a cloud sales-intelligence platform for revenue teams. Clearminutes is a local-first desktop app with no bots, no uploads, and on-device AI.
Best for: revenue teams that want conversation intelligence, CRM auto-entry, and AI coaching scorecards on top of meeting notes. A bot or integration joins every call and audio is uploaded to Avoma's cloud on AWS. There's no free tier, just a 14-day trial, and no on-device transcription.
Price: $19/seat/mo/annual
Avoma is best for revenue and customer-success teams who want conversation intelligence, CRM auto-entry, and AI coaching scorecards on top of meeting notes. A bot or integration joins every call and audio is uploaded to Avoma's cloud on AWS, with no free tier beyond a 14-day trial and no on-device transcription. Strong on Salesforce, HubSpot, and coaching analytics, weak on privacy and any offline or no-bot workflow. Pick it when revenue analytics matter more than keeping audio on the device.
Two local-first Tauri apps built on similar technology. Meetily is open source with a new Pro tier. Clearminutes adds speaker diarization, integrations, guided onboarding, and active commercial development.
Best for: developers who want an open-source local notes tool that runs Whisper on-device with unlimited free use. There are no integrations and no diarization polish. The split is Free Community / $10 Pro, and on Linux it's buildable from source rather than a packaged app.
Price: £0/forever
Meetily is best for developers who want a fully auditable, MIT-licensed local notes tool that runs Whisper on-device with unlimited free use. There are no integrations, diarization is only coming soon on Pro, and there's no commercial support, just the community. On Linux it's buildable from source rather than a packaged app. Strong on openness and local processing, weak on polish, integrations, and support. Pick it when an auditable open-source codebase matters more than a finished product.
Notion AI Meeting Notes lives in your Notion workspace and sends audio to cloud sub-processors. Clearminutes keeps audio on your machine with no account and no uploads.
Best for: teams already living in Notion who want meeting notes inline in their workspace. Audio goes to OpenAI, Anthropic, and other third-party LLMs, with no on-device transcription. Speaker identification is English-only and 1:1-best, and a Notion account is required.
Price: £0/forever
Notion AI Meeting Notes is best for teams already living in Notion who want meeting notes inline in their workspace. Audio goes to OpenAI, Anthropic, and other third-party LLMs, with no on-device transcription, and a Notion account is required. Speaker identification is English-only and works best 1:1. Strong on in-workspace notes and custom templates, weak on privacy, offline, and any no-account workflow. Pick it when workspace notes matter more than keeping audio off third-party servers.
Trilla is a free local-first Mac app on Apple Silicon. Clearminutes adds Windows, task sync, and a Pro tier.
Best for: Mac users on Apple Silicon who want a free, local, no-bot transcriber running Whisper and Ollama on-device, with local voice prints and no account. It's macOS-only, with no Windows or Linux build. The integration surface is smaller and the feature set is lighter than the cross-platform options.
Price: £0/forever
Trilla is best for Mac users on Apple Silicon who want a free, local, no-bot transcriber that runs Whisper and Ollama on-device with local voice prints and no account. Every feature is free with no subscription and no gating. It's macOS-only, with no Windows or Linux build, and it relies on Ollama for local summaries rather than shipping a built-in LLM. Strong on free local capture, weak on cross-platform and integrations. Pick it when free and local on Mac matters more than Windows or Pro integrations.
On-device, no bot, no cloud audio.
Best for: privacy-first teams who want meeting notes without uploading audio to the cloud.
Price: £0/forever
Clearminutes is best for teams who need meeting audio to stay on the device. Transcription runs on local Whisper and summaries on a built-in local LLM, with no cloud upload at any step. It works offline, ships for Windows alongside macOS, and shows zero outbound audio connections in its in-app network monitor. A real free tier and bring-your-own API keys make it the local-first pick when every other tool on this list sends audio to the cloud.
| Feature | Granola | Avoma | Jamie | Fireflies.ai | Krisp | tl;dv | Trilla | Otter.ai | Fathom | Meetily | Notion AI Meeting Notes | Clearminutes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Meeting bots | No | Yes (bot or integration joins call) | No | Yes (bot joins call) | No (bot-free option) | Yes (bot joins call) | No | Bot joins call | Yes (bot joins call) | No | No | No |
| Local processing | No (cloud-only) | No (cloud-only) | Partial (offline available) | No (cloud-only) | Partial (noise cancellation local, transcription cloud) | No (cloud-only) | Yes (Whisper + Ollama) | No (cloud-only) | No (cloud-only) | Yes (Whisper) | No | Yes (on-device models) |
| Speaker identification | Yes | Yes | Yes | — | Yes | Yes | Yes (local voice prints) | Yes | Yes | — | Yes (English only, 1:1 best) | Yes (Pro) |
| Works offline | — | No | — | No | Partial (noise cancellation offline) | No | — | No | No | — | No | Yes (local models) |
| Free tier | Yes (limited history) | No (14-day trial only; viewers free on paid plans) | Yes (limited) | Yes (limited AI summaries, 800 min storage) | Yes (14-day trial, then limited) | Yes (limited recordings) | Yes (unlimited) | Yes (300 min/month) | Yes (unlimited recordings) | Yes (unlimited, open source) | Limited AI trial (Free/Plus) | Yes |
| Language support | — | 40+ languages (translation on Organization tier+) | — | 69 languages | 16 languages | 30+ languages | — | English, French, Spanish | 28 languages | — | 16 languages | 6+ languages |
These aren't bugs. They're the shape of a Mac-only, post-stop, menu-bar-first tool, and each one bites a specific kind of buyer.
They bite only when your requirement is cross-platform, a live transcript, or network proof. If you live on a Mac, don't need a live view, and take the local-first claim on trust, none of these bite, and Speechmark's menu-bar plus MCP plus one-time-price shape is a clean pick.
This is a close match, local-first against local-first, and the difference is narrower than a cloud comparison. Both keep audio on the machine and skip the meeting bots. Speechmark optimises the menu-bar, MCP, and one-time-price shape, and accepts Mac-only, post-stop processing, and no monitor. Clearminutes optimises where the audio goes and what you can prove about it, and accepts a subscription on the Pro tier.
Your hard requirement is that the tool runs on Windows or Linux as well as macOS, that you see a live transcript while you record, or that an in-app privacy network monitor proves zero outbound connections during a meeting. Clearminutes captures microphone and system audio on the machine, transcribes locally, works offline, and shows the network monitor so the local-first claim is verifiable, not asserted. No bot joins the call. A built-in local LLM and bring-your-own API keys cover the notes step. For a mixed-OS team, a buyer who reads along mid-call, or anyone who needs to show a compliance reviewer the zero-connections proof, this is the structural difference. The trade is the price shape: Clearminutes is free for local use and £12 a month for Pro, where Speechmark charges a one-time $49 to $79 and nothing after. That's the right trade when the constraint is cross-platform, live, or proof, not the one-off price.
You're on a Mac, you want a menu-bar capture app with no account and no login, you like the one-time $49 to $79 price (free during beta, no credit card), and you already work through Claude Desktop so the MCP connector fits your flow. Speechmark keeps audio on the Mac, transcribes on Apple Foundation Models by default, supports a confirm-name speaker ID, and ships action items. If you don't need Windows or Linux, don't need a live transcript, and take the local-first claim on trust without a monitor, Speechmark is the cleaner, cheaper, one-off pick.
Do the price maths before you commit. Speechmark is free during beta and $49 to $79 once, per Mac, no subscription. Clearminutes is free for local use and £12 a month for Pro. For a solo Mac user who doesn't need cross-platform, live, or a monitor, the one-time price is cheaper the day you buy it and cheaper every year after. For a user who needs any of the three structural features, the subscription buys something the one-off price doesn't, and the comparison stops being about money.
Neither tool is the better pick across the board. Speechmark keeps the menu-bar flow, the Claude Desktop connector, the no-account simplicity, and the one-time price. It gives up Windows, Linux, the live view, and any in-app proof of the privacy claim. Clearminutes keeps cross-platform support, the live transcript, the network monitor, and offline use. It gives up the menu-bar simplicity and charges a monthly subscription on Pro. Pick the one whose trade-offs you can live with.
Last updated: 2026-07-01. Compared from each tool's public pricing and feature data (see sources), as of the last-updated date.
Sources: Speechmark official site (2026-06-21), Speechmark pricing page (2026-06-21), Speechmark privacy page (2026-06-21).
Clearminutes is our own product; we've kept the comparison fair.
Local transcription, live transcript view, AI summaries, and a verifiable privacy network monitor. No cloud uploads, no bots, runs on macOS, Windows, and Linux.