Narada vs Jamie — “no bot” is not the same thing as “on your Mac”
Both apps record your Mac’s own audio and neither one joins your call. The divergence is what happens in the second after you stop recording — Jamie uploads the audio to Germany, Narada transcribes it where it already is.
- Jamie is bot-free but not local. Its own documentation says “The audio recording from your computer is uploaded to a server location in Frankfurt, Germany”, processed on serverless GPU, after which “we delete all audio files permanently.”
- Narada transcribes on the Mac using Parakeet TDT v3 through Core ML. The audio never leaves the machine in any mode; in Pro AI or bring-your-own-key mode the transcript text goes to OpenAI after an explicit consent step.
- Jamie’s free plan is 10 meetings a month capped at 30 minutes each, and one credit is charged per meeting regardless of length. Plus is €25/month, Pro €47/month (€21 and €39 billed annually), plus VAT. Checked 4 August 2026.
- Jamie’s docs say it supports “over 80 languages”; its homepage says 99+. Narada transcribes 25 European languages, with the model vendor’s own word error rates ranging from 4.0% (Italian) to 36.9% (Greek).
- Jamie runs on macOS 13.1+, Windows and iOS. Narada is macOS 15.6 or later and nothing else — no Windows app, no iPhone app.
If you want meeting notes without a bot in the participant list, both of these will do it. Pick Jamie if you need Windows or an iPhone, more than 25 languages, or an ISO 27001 certificate for a procurement team. Pick Narada if the audio must not leave your Mac, if you record more meetings than a credit allowance covers, or if you want the transcript as a plain text file in a folder you chose.
Jamie is the closest thing to Narada in intent — a native desktop app, deliberately bot-free, sold on privacy. It is also the clearest illustration of where “bot-free” stops. Jamie’s audio goes to Frankfurt. Narada’s does not go anywhere.
Who each one is for
Jamie suits someone whose meetings span languages Narada’s speech model does not cover, who works across a Mac and a Windows machine, or who needs audited certifications on a page a compliance officer can read. If your objection to Otter and Fireflies was “American servers” rather than “servers”, Jamie answers it directly.
Narada suits someone whose objection was the second one. It records your microphone and your Mac’s output as two separate tracks, transcribes both on the machine, and writes .m4a, .txt, .md and .json into a folder you pick. No account needed to record, no per-meeting metering, no web app. It is also narrower: no Windows, no phone, no team workspace, no CRM integrations.
The difference that actually matters
Jamie documents its pipeline honestly, which makes this easy to state. From its data FAQ:
“The audio recording from your computer is uploaded to a server location in Frankfurt, Germany, and then processed on a serverless GPU environment from Modal (no data is stored at Modal). Once the processing is done and your transcript is ready, we delete all audio files permanently."
And then:
“After the generation of your transcript, it is further processed via an Anthropic or OpenAI API to generate your meeting notes.”
So: capture locally, upload to Germany, transcribe on someone else’s GPU, generate notes through Anthropic or OpenAI, store the text in the EU, delete the audio. Every step is defensible and Jamie defends them well. But the recording of your voice and your client’s voice does travel over the network, and there is a window in which it exists on hardware you do not own.
Narada has no upload step for audio at all. Transcription runs through Core ML on your Mac using Parakeet TDT v3 — a roughly 600 MB model downloaded once — and speaker separation runs locally on the call’s track. Choose On-device AI for summaries and nothing leaves the machine at any point. Choose Pro AI or your own OpenAI key and the transcript text is sent to OpenAI after you accept a consent sheet that says so. The audio file stays put in every mode.
That is the whole trade. “Processed in the EU under ISO 27001” and “processed on your Mac” are different guarantees, and which you need depends on whether your constraint is jurisdiction or transmission.
Side by side
| Jamie | Narada | |
|---|---|---|
| Bot in the participant list | No — “The AI note taker. Without a bot.” | No |
| How audio is captured | Native desktop app; “monitors your computer’s microphone status to detect when you’re in a meeting, captures system audio” | CoreAudio process tap on system output + microphone, recorded as two separate tracks |
| Where transcription happens | Uploaded to a server in Frankfurt, transcribed on Modal’s serverless GPU | On the Mac, Parakeet TDT v3 via Core ML |
| Where summaries are generated | “via an Anthropic or OpenAI API” | Your choice of five: nothing, on-device (Apple Silicon), Pro AI, your own OpenAI key, or your own endpoint |
| Does the audio leave your machine | Yes, then “we delete all audio files permanently” | Never, in any mode |
| Works with no internet | No — transcription is server-side | Yes for recording and transcription; also for summaries in On-device AI |
| Transcription languages | Docs say “over 80 languages”; homepage says 99+, with 13 flagged as variable quality | 25 European languages; model vendor’s word error rates run 4.0% (Italian) to 36.9% (Greek) |
| Speaker attribution | Not specified on the vendor pages checked | Your mic track is named; the other side is split into unnamed Speaker 1, Speaker 2 |
| Per-meeting limits | 1 credit per meeting regardless of length; 30 min (Free), 2 h (Plus), 3 h (Pro/Team) | None. Disk space is the limit |
| Where your archive lives | Transcript and summary on Jamie’s server in Frankfurt | A folder you choose on your Mac, as unencrypted plain files |
| Platforms | macOS 13.1+, Windows, iOS | macOS 15.6+, universal binary. On-device AI needs Apple Silicon |
| Certifications | GDPR, ISO 27001, DORA; “SOC 2 Type II (pending)" per its own pricing table | None claimed |
| Entry price | Free: 10 meetings/month, 30 min each | Free: complete recorder, unlimited meetings |
| Paid | Plus €25/mo, Pro €47/mo (€21 / €39 billed annually), plus VAT | Pro AI $6.99/mo or $69.99/yr, 3-day trial |
Prices and vendor quotes checked 4 August 2026. Verify before buying.
Where Jamie is the better choice
If you take meetings in Japanese, Arabic, Hindi, Turkish, Korean, Chinese or Norwegian, Narada cannot transcribe them. Parakeet v3 covers 25 European languages and that list is the list; Jamie’s documentation claims over 80. For a multilingual sales team this is not a close call.
If you use a Windows machine at work and a Mac at home, or want notes from in-person conversations on your phone, Jamie has apps for that and Narada has none. Jamie also runs on macOS 13.1, where Narada requires 15.6 — on Ventura or Sonoma, Narada is not available to you at all.
And if someone in legal asked which certifications the vendor holds, Jamie has answers and Narada does not: ISO 27001 with an independent audit, DORA, a DPA, and “Your data is never used for the training of models, not by us or any of the model providers.” Narada’s answer is that there is no vendor-side copy of your meetings to certify — true, and much harder to put in a procurement form.
One more. Jamie’s summaries come from frontier cloud models; Narada’s On-device AI is worse than that, and the app says so in its own settings — “lighter quality than Pro” — capping on-device prompts at 24,000 characters against 100,000 for cloud. If summary quality is the whole point and local processing means nothing to you, Jamie’s notes will read better.
Where Narada is the better choice
Anything where transmitting the recording is itself the problem. A therapist’s session, a lawyer’s client call, a journalist’s source, an internal investigation, an NDA that names transfers rather than jurisdictions. “Uploaded to Frankfurt and deleted afterwards” is a good policy and it is still a transfer. On-device transcription means there is nothing to transfer and nothing to delete.
Volume, too. Jamie charges one credit per meeting regardless of duration, so a four-minute stand-up costs the same as a ninety-minute review, and Plus gives you 20 a month. Eight short calls a day burns that in a week and lands you on Pro at €47/month. Narada has no per-meeting accounting because there is no server counting.
The archive matters more than people expect. Narada writes one folder per meeting: audio, the merged transcript with timestamps, the raw microphone and system transcripts, a Markdown summary, the analysis as JSON. Deliberately unencrypted, so you can grep it or feed it to something else, and readable even if Narada disappeared tomorrow. There is also a Claude Desktop connector — six read-only tools that search your transcripts and follow-ups locally, though the matching is keyword, not semantic.
Two other bot-free options, and how their mechanics differ
Bluedot captures through a Chrome extension or a desktop app: “Bluedot never invites a bot to your meeting or asks for calendar access.” Its settings offer Cloud, Local and Auto processing — but by the vendor’s own description that choice governs “where your recording will be stored while it is being processed", not whether the meeting stays on your Mac. Its privacy policy lists Google (Alphabet Inc.) receiving “video transcriptions”, “Place of processing: United States”.
Krisp does real on-device inference, but installs a virtual audio device to get it: “Krisp Microphone and Krisp Speaker are virtual devices… They do not operate with other virtual microphones”, so it collides with BlackHole or Loopback. On-device transcription is English only — the other 15 languages are “server-based”. Summaries always leave, via “Microsoft Azure services”, and all stored data is in the US.
Both also sell a bot mode, so for them “no bot” is the default rather than the architecture.
Privacy and where your audio goes
Both vendors state their position clearly. Compare them in their own words.
Jamie, on residency:
“All data storage and processing stays within the EEA, Switzerland, and the UK. No data leaves these jurisdictions, ever.”
On training:
“Data is not stored at Anthropic or OpenAI, and not used for training purposes.”
Narada, from its own consent sheet and settings: in Recorder-only mode, “This Mac only. No AI requests are sent.” In On-device AI, nothing leaves after the one-time model download. In Pro AI or bring-your-own-key, transcript text and chat questions go to OpenAI — via Narada’s proxy or directly with your key — and the consent text says “Narada uses the OpenAI API (not ChatGPT). OpenAI does not use API inputs to train its models by default.” That is a statement about OpenAI’s policy, not a guarantee Narada can make on OpenAI’s behalf.
Two caveats on Narada’s paid tier, because the app’s onboarding says it does not phone home and that is only true in the two local modes. In Pro AI, the app sends a per-install device identifier and a subscription identifier to Narada’s own proxy, and stamps each AI request with a request kind and a hashed meeting id so server costs can be attributed. No product-analytics events are sent, but the proxy is a server with a database. And the files are unencrypted on purpose — a convenience choice, not a security one. If your threat model includes someone with access to your Mac, full-disk encryption is your protection, not the app.
Neither app makes recording lawful for you. Narada keeps a “notify participants” reminder on screen for the whole recording; Jamie’s docs describe a prompt when it detects a call. The consent conversation is yours to have.
Pricing, plainly
Jamie’s free plan is a real free plan: €0, 10 meetings a month, 30-minute cap, no credit card. The catch is at the boundary — Jamie’s own docs say that once credits or the duration limit are exceeded, “your meeting notes will be locked, and you will need to upgrade to a paid plan to access them.” Notes you already recorded, behind the paywall. Paid: Plus €25/month (€21 annually), Pro €47/month (€39 annually), Team €39 per member (€33 annually), all “Plus applicable VAT”, Enterprise custom.
Narada’s free tier is the whole recorder: calendar-aware recording when you enable it, microphone and system capture, on-device transcription with speaker separation, plain files, labels and notes, storage management, the integrity checks and the Claude connector. The $6.99/month or $69.99/year Pro AI subscription buys cloud summaries, follow-ups, chat over the archive and pre-meeting briefs — all of which are also free via On-device AI, which needs Apple Silicon, 16 GB of RAM for the larger model and a one-time 3 GB or 7 GB download. Both Pro plans have a 3-day trial.
How to try both
Run them in parallel for a week. They do not fight: Jamie prompts when it detects a call, Narada taps your Mac’s output through CoreAudio, and neither installs a virtual audio device that could conflict with the other. Narada asks for Microphone and System Audio Recording and never for Screen Recording.
The test that settles it is one meeting you would be uncomfortable uploading. If knowing the audio went to Frankfurt and came back deleted bothers you, that is your answer. If you finish the week annoyed instead that Narada missed a Greek phrase or has no iPhone app, that is your answer too. And if you leave Jamie, export the transcripts before you cancel — the notes live on their server.
Sources
- https://meetjamie.ai — “The AI note taker. Without a bot.”; languages.
- https://meetjamie.ai/security — EEA/Switzerland/UK residency, encryption, ISO 27001, DORA.
- https://meetjamie.ai/pricing — both billing cycles, VAT note, “SOC 2 Type II (pending)".
- https://meetjamie.ai/blog/llm-info — capture mechanism, platforms, monthly prices.
- https://docs.meetjamie.ai/faqs-troubleshooting/data — Frankfurt upload, Modal GPU, Anthropic/OpenAI, audio deletion, GDPR.
- https://docs.meetjamie.ai/enterprise/security-and-compliance — EU-only processing, no model training.
- https://docs.meetjamie.ai/faqs-troubleshooting/language-support — “over 80 languages”.
- https://docs.meetjamie.ai/billing-pricing/subscriptions-plans — credit model, prices, notes locked over the limit.
- https://docs.meetjamie.ai/getting-started/recorder — how recording is started.
- https://bluedothq.com and https://bluedothq.com/privacy, plus https://help.bluedothq.com on recording settings and the desktop-app FAQ.
- https://krisp.ai/security-for-ai-meeting-assistant, https://krisp.ai/ai-meeting-assistant, and https://help.krisp.ai on the virtual devices.
- FluidInference/FluidAudio docs — Parakeet TDT v3's 25 European languages and FLEURS word error rates. https://raw.githubusercontent.com/FluidInference/FluidAudio/main/Documentation/Benchmarks.md
All checked 4 August 2026. Narada’s figures come from its shipping build (1.2, minimum macOS 15.6) and its in-app text.
Questions people ask about this
Does Jamie send my meeting audio to the cloud?
Yes. Jamie’s data FAQ states the recording “is uploaded to a server location in Frankfurt, Germany” and processed on a serverless GPU environment from Modal, and that the audio files are deleted permanently once the transcript is ready. The transcript and summary are then stored in Frankfurt.
Is Jamie GDPR compliant?
Jamie is a German company and states on its security page that “All data storage and processing stays within the EEA, Switzerland, and the UK. No data leaves these jurisdictions, ever.” It lists GDPR, ISO 27001 and DORA, and its own pricing table lists SOC 2 Type II as “pending”. Narada makes no compliance claim of any kind — it has no server holding your meetings to make a claim about.
Which one works without an internet connection?
Narada. Recording and transcription are local, and with On-device AI selected the summaries are local too. Jamie cannot produce a transcript offline because transcription is server-side.
Does Narada name the speakers?
No. Your own voice is labelled because your microphone is recorded as a separate track. Everyone else is split into “Speaker 1”, “Speaker 2” and so on — the app has no way to know their names and does not guess.
Try it on your next call
Narada records your Mac's audio and your microphone, transcribes on-device, and hands you the decisions and follow-ups. Free to use, no account, no bot in the meeting.