tl;dv is Berlin-based, has a free plan, a recording library with clips, support for dozens of languages and thousands of integrations. As a meeting notetaker it is good and cheap: Pro is USD 18 per seat per month billed annually, Business is USD 59 (pricing as of August 2026). So this comparison only makes sense if your problem does not end at the note. If it does — take tl;dv and stop reading.
Where the note ends and the work begins
After the meeting somebody has to move the stage, update the amount, write the follow-up, create the tasks and correct the quarterly forecast. tl;dv pushes the note out and maps fields into HubSpot, Salesforce or Pipedrive. The rest stays with the rep — which is to say, with their evening.
In Luminote the meeting is an input to the system, not its output:
- What was agreed fills in the deal fields, and decisions from the call turn into tasks with an owner and a due date.
- Signals from the call feed an engine that scores deal health from 0 to 10 — with time decay, so a signal from two months ago weighs less than yesterday's.
- The forecast is signed off by the rep and kept with its change history. After the quarter you can reconstruct who promised what and when they changed their mind.
- Automations run on fourteen triggers: a silent champion, a MEDDIC threshold, a quote opened twice with no reply, a deal that has not moved. With a queue, retries and a view of whatever failed — plus a mode where the system prepares and a human approves.
There is one more difference, unspectacular but useful in practice. Lumi carries out instructions rather than only answering them: “note that the customer wants a quote by Friday” ends as a task in the CRM. And questions about numbers — pipeline value, deals at risk — go straight to the database rather than to the model. The model cannot get an amount wrong, because we never ask it for one.
Analysis: one meeting versus sixteen kinds of meeting
tl;dv treats every meeting the same way: recording, transcript, summary, optionally a scorecard from a playbook.
Luminote recognises 16 meeting types and scores each against a different rubric: discovery, presentation, negotiation, decision, checkpoint, onboarding, QBR, partner, transactional, account, internal, project, retrospective, recruitment, supplier meeting — and “other”. A team retrospective does not get marked down for “champion engagement”, because it was not a sales call, and the system knows that. The classifier proposes the type and the user can change it.
Privacy: EU hosting versus EU hosting plus anonymisation
In fairness: tl;dv does more here than most of its American competition. It hosts data in Europe, transcribes with Whisper, holds SOC 2 and states that it does not train models on customer data. But AI processing can take a European or an American path, and the core of the product — the recording library — depends by definition on transcripts and video staying put.
Our arithmetic is different, because the priorities are different:
- Anonymisation before the model — the model sees “[Person 1]”, not your customer's name.
- The transcript is deleted after analysis — what remains is what follows from it. Keeping the recording has to be switched on deliberately.
- No OpenAI. Analysis runs on Mistral in Paris; transcription runs on Whisper locally on the device where the hardware allows, or on Voxtral in Paris.
- All of production in the European Union, with a public subprocessor list and thirty days' notice before it changes.
This is a deliberate trade: we give up part of what a recording library does in exchange for the raw record of your conversations not accumulating here. If your legal team would rather have it the other way round, tl;dv is closer to your preference.
CRM: pushing a note versus syncing both ways
tl;dv pushes notes and maps fields into three large CRMs. Luminote has 24 connectors: HubSpot, Salesforce, Pipedrive, Zoho, Close, Teamleader, Sellsy — and, which nobody else in this class carries, Livespace, Firmao and Raynet.
The sync runs both ways across deals, contacts, companies and tasks — with rollback, a connection health monitor and a permissions filter, so a rep never sees deals in Luminote that they cannot see in their own CRM. Field mapping proposes itself: you approve it rather than fill it in.
Side by side
| Luminote | tl;dv | |
|---|---|---|
| Category | Sales platform: pipeline, forecast, quotes, inbox, automations | Meeting notetaker with coaching |
| Entry price | 14-day trial, no card; Basic for the team, Pro for AI users | free plan with an AI-summary limit; Pro USD 18, Business USD 59 per seat |
| Hosting | European Union, no transfer outside the EEA | Europe |
| AI processing | Mistral in Paris, no OpenAI | Whisper and OpenAI; European or American path |
| Anonymisation before the model | ✅ | — |
| Transcript retention | deleted after analysis; recording only on request | a library of recordings and transcripts — the core of the product |
| Meeting types | 16, each with its own rubric | uniform analysis plus playbook scoring |
| CRM | 24 connectors, two-way sync, permissions filter | HubSpot, Salesforce, Pipedrive — note and field push |
| Forecast | signed off, with change history | ✗ |
| Automations | 14 triggers based on buyer behaviour | ✗ |
| Interface | Polish, English, German | English, German and others |
When to pick tl;dv, and when Luminote
tl;dv is the right call when what you need is recordings and notes, the budget is tight, and a library of conversations with clips you can share is a value in itself. The tl;dv free plan does things other vendors charge for.
Luminote makes sense when the notes are not the problem but the symptom. The problem is a CRM nobody updates, a forecast assembled in a spreadsheet on the last Friday of the quarter, follow-ups written after 6pm — and a legal team asking what exactly the model saw and how long you keep it.
Key takeaway
If all you need after a meeting is a note, tl;dv will do it more cheaply than we will. If after every meeting somebody still has to move the stage, write the email and correct the forecast, then what you are buying is not a notetaker but the system that does it.