You want your app to capture what happens in a meeting, and the vendor market hands you three overlapping product categories that all claim to do it: recorders, transcription services, and notetakers. They sound interchangeable in marketing copy, but they solve different problems and produce different outputs. Pick the wrong one and you either ship a feature that’s missing the data you needed, or you bolt three separate services together and maintain three integrations.
This guide draws the line between a meeting recorder, a transcription API, and a notetaker, shows where each fits, and explains why the combined Notetaker API usually replaces all three for in-meeting capture across Zoom, Google Meet, and Microsoft Teams.
What is the difference between a meeting recorder and a transcription?
Section titled “What is the difference between a meeting recorder and a transcription?”A meeting recorder captures the raw audio and video of a call and writes it to a file, typically an MP4 or audio stream. A transcription service takes that audio and converts speech into time-stamped text. The recorder gives you the media; the transcription gives you the searchable words you can index and analyze.
The split matters because each output drives different features. A recording is what you keep for compliance, playback, or training review, and it’s heavy: a one-hour call can land around 50 MB to several hundred MB depending on resolution. A transcript is light, usually a few hundred kilobytes of JSON, and it’s what powers search, sentiment, and AI summaries. Most teams discover they need both, which is why running a standalone recorder and a separate speech-to-text vendor means stitching two pipelines together and reconciling timestamps between them.
Recorder vs transcription vs notetaker
Section titled “Recorder vs transcription vs notetaker”A notetaker combines a recorder and a transcription service into one workflow: a bot joins the meeting, captures audio and video, transcribes the speech, and generates a summary and action items from the transcript. Instead of running three vendors, you make one call and receive every output keyed to the same meeting. The table below maps each category to what it actually produces.
| Capability | Recorder only | Transcription only | Nylas Notetaker API |
|---|---|---|---|
| Captures audio/video | Yes (MP4) | No | Yes, video_recording and audio_recording |
| Speech-to-text | No | Yes (text) | Yes, transcription |
| Joins the call as a bot | Sometimes | Rarely | Yes, across Zoom, Meet, Teams |
| AI summary | No | No | Yes, summary |
| Action items | No | No | Yes, action_items |
| Integrations to maintain | 1 | 1 | 1 |
A pure recorder skips the text entirely, so you’d pair it with a speech-to-text API and write glue code to align the two. A pure transcription API expects you to feed it audio it never captured, so you still need something inside the meeting. The combined approach collapses both jobs, plus the AI layer on top, into a single grant-scoped request.
What is a meeting AI assistant and how does it connect to meetings?
Section titled “What is a meeting AI assistant and how does it connect to meetings?”A meeting AI assistant is a bot that joins a video call, records it, transcribes the speech, and produces structured notes such as a summary and action items. It connects through the meeting URL: you pass the Zoom, Google Meet, or Microsoft Teams link, and the bot dials in as a participant.
The Notetaker API exposes this through POST /v3/grants/{grant_id}/notetakers. You send a meeting_link, an optional join_time as a Unix timestamp so the bot arrives on schedule, and a meeting_settings object that switches each capability on. You can also schedule the bot against a connected calendar so it auto-joins events. One detail to plan around: the API doesn’t de-duplicate bots, so every create request invites a new Notetaker to that meeting. The request below sends a bot with recording, transcription, summary, and action items all enabled.
curl --request POST \ --url 'https://api.us.nylas.com/v3/grants/<NYLAS_GRANT_ID>/notetakers' \ --header 'Authorization: Bearer <NYLAS_API_KEY>' \ --header 'Content-Type: application/json' \ --data '{ "meeting_link": "https://meet.google.com/xyz-abcd-ijk", "name": "Nylas Notetaker", "meeting_settings": { "video_recording": true, "audio_recording": true, "transcription": true, "summary": true, "action_items": true } }'What is an API for recording meetings automatically?
Section titled “What is an API for recording meetings automatically?”An API for recording meetings automatically sends a bot to a call without a human pressing record, then exposes the captured files programmatically. You enable recording in the create request, and once the bot leaves, the media becomes available to fetch over HTTP. The same request that records also transcribes, so you don’t run a second job to get text.
Retrieval happens at GET /v3/grants/{grant_id}/notetakers/{notetaker_id}/media. The response returns one object per enabled capability, recording, transcript, summary, and action_items, each with a pre-authenticated url. Those URLs are valid for 60 minutes from generation, so you have a one-hour window to download the bytes onto your own storage before the link stops working. The call below pulls the media manifest. Fetch it as soon as the recording-available webhook fires, then copy each file before the hour runs out.
curl --request GET \ --url 'https://api.us.nylas.com/v3/grants/<NYLAS_GRANT_ID>/notetakers/<NOTETAKER_ID>/media' \ --header 'Authorization: Bearer <NYLAS_API_KEY>'When should you use a standalone recorder or transcription service?
Section titled “When should you use a standalone recorder or transcription service?”Use a standalone recorder or a dedicated transcription API when you control the audio source and don’t need a bot in the call. If your app already streams a podcast file, a phone-system recording, or an uploaded MP3, you only need speech-to-text, so a meeting bot adds overhead. Recorders that hook into a Zoom cloud recording also skip the bot model.
That tradeoff is real and worth respecting: the bot approach shines when you’re capturing live multi-party video calls across providers you don’t own, and it’s overkill when you have a single audio file in hand. The break point is provider spread. The moment you need to record Zoom, Google Meet, and Microsoft Teams from one code path, three separate integrations turn into one. The bot model adds a few moving parts you don’t get with a raw file: the bot moves through state values such as scheduled, attending, and available, and you can set leave_after_silence_seconds so it drops off an empty call instead of recording dead air. For the full setup, see the meeting notetaker API guide, and compare bot vendors in the Recall.ai alternative guide.
What’s next
Section titled “What’s next”- Meeting notetaker API guide to send your first bot to Zoom, Meet, or Teams
- Get a meeting transcript and recording for downloading media before the links expire
- Get notified when a transcript is ready to trigger fetches from webhooks
- Recall.ai alternative to compare meeting bot vendors