Skip to content

Voice Notes

Voice notes let you capture thoughts, meetings, or spoken content by recording audio. Atlas handles the rest — transcription, summarization, and saving to your vault as a structured note.

Recording a voice note works in three steps:

Open the voice note recorder (the waveform icon in the toolbar). Press Record and start speaking. You can record for any length of time — a quick 30-second thought or a 45-minute meeting.

When you’re done, press Stop.

After you stop recording, Atlas processes the audio:

  • The recording is split into chunks (roughly 2 minutes each) for reliable transcription
  • Each chunk is sent to OpenAI Whisper via the Atlas cloud for transcription
  • The transcription chunks are merged into a single, continuous transcript
  • Atlas generates an AI summary with key topics and action items
  • A quality report is included showing confidence scores for the transcription

Long recordings take longer to process. A 10-minute recording might take 20–30 seconds. You’ll see a progress indicator while it works.

Before anything is saved, you see the full results:

  • Transcript — the complete text of what was said
  • AI summary — key points, topics, and any action items Atlas identified
  • Quality report — confidence indicators that flag any parts of the transcription that may be inaccurate

You can edit the transcript directly before saving. If a word was misheard or a proper noun was wrong, fix it here. The summary is regenerated from your edits.

When you’re happy with the result, click Save to Vault. The note is saved to your vault with structured frontmatter including the date, duration, and topics.

Voice notes are saved as Markdown files in your vault. Each note includes:

---
type: voice-note
date: 2026-02-24
duration: 8m 14s
topics: [meeting, Q2 planning, action items]
---
## Summary
...
## Transcript
...

You can find them in your vault like any other note, search them with Atlas, or link them from other notes.

Voice notes also support phone call recording mode with speaker diarization. When you record a two-person conversation, Atlas can distinguish between the two speakers and label their turns in the transcript as Speaker 1 and Speaker 2.

To use this mode, select Phone Call in the recorder before starting. This mode works best with clear audio where the two speakers don’t talk over each other.


Next: How Memory Works