> ## Documentation Index
> Fetch the complete documentation index at: https://docs.vowen.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Summaries

> Read, edit, regenerate and fix the AI summary of a meeting note.

When a recording ends, Vowen transcribes it and sends the transcript to your AI provider with your [template](https://docs.vowen.ai/meeting-notes/templates.md). The result is the note's summary: a title and the sections the template asked for. You can edit it, regenerate it with another model or template, and ask the AI to rewrite it.

Summaries need a language model. Set one up under **Language models** in the sidebar. See [AI setup](https://docs.vowen.ai/ai-features/ai-setup.md).

![A meeting note on the Summary tab with citation pills, action-item checkboxes and the Details inspector on the right](https://mintcdn.com/vowen/jjhC7JNGyK5c2anS/images/screens/note-summary.webp?fit=max\&auto=format\&n=jjhC7JNGyK5c2anS\&q=85\&s=23dd3d567a06bf7f1a45cad8ac9e2c17)

## What the summary is built from

Vowen sends the AI:

* The transcript, with speaker names where you have them.
* Anything you typed in the meeting panel's **Notes** tab or the note's **My Notes** tab.
* Moments you marked as important, with an instruction to weigh them more.
* Screenshots you captured during the meeting, when the model can read images.
* Your template's prompt.

Vowen writes the note's title itself. You do not need to ask for one in your template.

## The note page

Open a note from **Notes**. The page has the audio player at the top, the title, and a row of chips under it: the date, the people on the call, and the note's tags. Click **+ add tag** to tag the note.

The tabs are:

| Tab | Shortcut | What it shows |
| - | - | - |
| **Summary** | <kbd>Cmd</kbd> + <kbd>1</kbd> / <kbd>Ctrl</kbd> + <kbd>1</kbd> | The AI summary, editable |
| **Transcript** | <kbd>Cmd</kbd> + <kbd>2</kbd> / <kbd>Ctrl</kbd> + <kbd>2</kbd> | The transcript with speakers and timestamps |
| **My Notes** | None | What you typed during the meeting. Only shown when there is something there. |
| **Ask AI** | <kbd>Cmd</kbd> + <kbd>3</kbd> / <kbd>Ctrl</kbd> + <kbd>3</kbd> | [Questions about the meeting](https://docs.vowen.ai/meeting-notes/ask-ai.md) |

The tab shortcuts use the physical number keys, so they also work on layouts such as AZERTY.

The panel on the right is the inspector. It has three tabs:

* **Details**: regenerate the transcript or summary, and send the note to a [webhook](https://docs.vowen.ai/integrations/webhooks.md).
* **Speakers**: name the speakers. Shown when the transcript has speaker labels. See [Speaker identification](https://docs.vowen.ai/meeting-notes/diarization.md).
* **Chats**: saved Ask AI conversations. See [Ask AI](https://docs.vowen.ai/meeting-notes/ask-ai.md#saved-conversations).

The **More options** (…) menu at the top holds **Open Recording**, **Show in Finder**, **Export**, **Copy**, **Send to webhook**, **Sync with Obsidian** and **Delete**, depending on what you have set up. The **Share** button next to it publishes the note as a link or saves it as a file. See [Share and export](https://docs.vowen.ai/meeting-notes/share-and-export.md).

![The Transcript tab with named speakers and timestamps](https://mintcdn.com/vowen/jjhC7JNGyK5c2anS/images/screens/note-transcript.webp?fit=max\&auto=format\&n=jjhC7JNGyK5c2anS\&q=85\&s=a83ee1ae961a4b6ac24a8f132c35754e)

## Edit the summary

Click into the summary and type. Changes save automatically.

* Type <kbd>/</kbd> to open the block menu: **Text**, **Heading 1**, **Heading 2**, **Heading 3**, **Bulleted list**, **Numbered list**, **To-do list**, **Quote**, **Code block**, **Divider** and **Image**.
* Insert an image by URL, or pick, drag or paste a file. Vowen copies local images into its own storage, so they stay with the note.
* To have the AI rewrite the summary or the title, ask on the [Ask AI](https://docs.vowen.ai/meeting-notes/ask-ai.md#edit-the-note-by-asking) tab.

### Find and replace

The summary and the transcript share one find bar.

* <kbd>Cmd</kbd> + <kbd>F</kbd> / <kbd>Ctrl</kbd> + <kbd>F</kbd> opens find.
* <kbd>Cmd</kbd> + <kbd>Shift</kbd> + <kbd>F</kbd> / <kbd>Ctrl</kbd> + <kbd>Shift</kbd> + <kbd>F</kbd> opens it with the replace row showing.
* **Replace** (<kbd>Enter</kbd>) replaces the current match. **All** (<kbd>Shift</kbd> + <kbd>Enter</kbd>) replaces every match in one undo step.

### Edit the transcript

Each transcript line has a **Segment actions** menu:

* **Reassign this line to** another speaker, or a **New speaker**.
* **Split at caret** breaks the line where your cursor is.
* **Merge with line above**.
* **Delete segment**.

Each of these is one undo step.

## My Notes

The **My Notes** tab holds what you typed during the meeting. It stays editable. **Regenerate summary from notes** saves your edits, regenerates the summary with them as context, and takes you back to the **Summary** tab. It needs a transcript.

![The My Notes tab with notes typed during the meeting and the Regenerate summary from notes button](https://mintcdn.com/vowen/jjhC7JNGyK5c2anS/images/screens/note-my-notes.webp?fit=max\&auto=format\&n=jjhC7JNGyK5c2anS\&q=85\&s=d0ea0d7c411ccc38bf09e3b2c0895c1d)

## Regenerate the summary or transcript

Use the inspector's **Details** tab.

**1. Open the note**

Click it in **Notes**. The inspector opens on **Details**.

**2. Pick a model and template**

In the **Summary** section, choose a model from **Model**, then pick a template from the list.

**3. Click Regenerate Summary**

The new summary replaces the old one. On a note with no summary yet the button reads **Generate Summary**.

The **Transcript** section above it works the same way. Choose a **Model**, set **Identify Speakers** and **Number of Speakers**, then click **Regenerate Transcript**. This re-transcribes the saved audio, so it only works while the recording exists. Without audio the section reads **Audio recording not available**.

The summary model list includes OpenAI, Anthropic, Groq, Gemini, DeepSeek, OpenRouter, Azure, Cerebras, Sarvam, Mistral, AWS Bedrock and Custom AI, plus Claude Code and Codex when they are installed. Only providers you have connected under **Language models** can be picked.

### Claude Code and Codex

If the Claude Code or Codex CLI is installed and signed in on your computer, Vowen lists it as a summary provider, labelled **Your CLI**. There is no API key to add. Claude Code runs with its tools turned off. Codex runs in a read-only sandbox. Both receive text only, so screenshots are not sent to them.

## Summary citations

Summary citations are a Pro feature and are off by default. Turn on **Summary citations** in **Notes › Notes settings › Recording Defaults**.

With citations on, the AI cites the transcript as it writes. Each citation is a small timestamp pill. Click one to open the **Transcript** tab at that line and cue the audio there. Playback does not start on its own.

* At most one citation appears per line.
* If the transcript is too long for the model to cite, the summary is written without pills.
* If you edit the transcript so much that a citation no longer matches, clicking it says the passage could not be found.
* A template can opt out. The **Follow-Up Email Draft** template does.
* When you export, citations can be removed or turned into plain `[12:04]` timestamps.

## Speaker names in summaries

When the transcript has speaker names, the summary uses them, so action items can name their owners. Name speakers on the inspector's **Speakers** tab and regenerate the summary to use new names. See [Speaker identification](https://docs.vowen.ai/meeting-notes/diarization.md).

With [voice profiles](https://docs.vowen.ai/meeting-notes/voice-profiles.md) on, Vowen waits up to 45 seconds for recognized names before it writes the summary.

## Screenshots and text-only models

If the summary model cannot read images, screenshots from the meeting are left out. The note shows "{count} screenshots weren't used in this summary" with a button to **Regenerate with** a model that can, or **Connect a model that can read images** if you have none. The inspector's **Summary** section shows **Text only: screenshots are skipped** for such models.

## When a summary fails

Vowen retries before giving up:

* After a recording, it makes 3 attempts, waiting 2 seconds and then 4 seconds between them.
* When you regenerate by hand, it makes 2 attempts with a 2-second wait.

It does not retry permanent errors: HTTP 400, 401, 403, 404, 413 or 422, an unconfigured provider, or a Claude Code or Codex CLI that is missing, signed out or fails to start. Rate limits and server errors are retried.

If every attempt fails, the note opens on the **Summary** tab with "Summary failed to generate. You can retry below." Click **What went wrong?** for the error, then pick another model and click **Generate Summary**. The transcript and audio are kept.

![A note whose summary failed, with the error disclosure and the Generate Summary button](https://mintcdn.com/vowen/jjhC7JNGyK5c2anS/images/screens/note-summary-failed.webp?fit=max\&auto=format\&n=jjhC7JNGyK5c2anS\&q=85\&s=603a13b131914bf034550a673cf36eba)

If the template has **Generate AI summary** turned off, the Summary tab reads "Summary is off for this template — only the transcript was saved." You can still generate one from the inspector.

## The audio player

* Playback speed from 0.5× to 2×. Press <kbd>\<</kbd> and <kbd>></kbd> to step. Pitch is preserved, and your speed is remembered.
* Skip back and forward 10 seconds.
* Clicking a transcript line or a citation pill cues the player to that moment.

## Where notes are stored

Notes are stored on your computer:

* **macOS:** `~/Library/Application Support/Vowen/`
* **Windows:** `%APPDATA%\Vowen\`

The audio is saved in your recordings folder, set in **Settings › Recording › Recordings folder**. See [Recording](https://docs.vowen.ai/meeting-notes/recording.md).

Deleting a note deletes its screenshots, its Ask AI conversations and the separate microphone and system-audio track files. The main recording stays in your recordings folder until you delete it yourself.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.