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

# Transcript, speakers and recordings

> Name the speakers in a meeting, play back the recording and decide how long Echo keeps audio.

Every meeting keeps its transcript. Open a meeting and select **Transcript** to read it, copy it with **Copy**, or play the recording if you kept one.

If you used [live translation](https://docs.callecho.ai/meetings/live-translation.md), the saved transcript keeps the original content. Translated captions are only available during the meeting and aren't saved.

## Name the speakers

Echo labels your microphone as you in online meetings. In some meeting apps, Echo can also see who the app shows as speaking; see [Supported platforms](https://docs.callecho.ai/reference/platforms.md). When that keeps pointing to the same person for a voice, Echo uses the name for all of that voice's lines, even ones where the app didn't show who was talking. Other voices get a number until someone names them.

After the meeting, a note above the AI Notes and the transcript says how many voices need review. Open it to see each voice with a line of what it said:

| The note shows | What it means |
| - | - |
| A numbered voice, such as Speaker 3 | Echo had no name to go on |
| **Suggested name** | The meeting app pointed to one person, but not often enough for Echo to use the name |
| **Conflicting names** | The meeting app showed different people for this voice. Open the naming options to review the competing names |

Echo doesn't apply a suggested or conflicting name by itself. You decide.

![The speaker naming editor open on an unnamed speaker from the Halden intake, with a sample line, a name field and the Save name, Leave unnamed and Play sample buttons.](https://mintcdn.com/callecho/3msX5gJvFFZ4hlDx/images/meetings/speakers-v1.png?fit=max\&auto=format\&n=3msX5gJvFFZ4hlDx\&q=85\&s=2661e2945b00ba8d3cf5d856d0d2f77f)

The screenshot shows an earlier editor layout. In the current editor, the name and excerpt appear first, with the naming options under a collapsed segment count.

1. Open the speaker review count and select **Name** or **Rename**, or select a speaker label in the transcript.
2. Type a name, or expand the naming options and use **Use a participant** to pick a meeting attendee. **Listen** plays an excerpt when its recording is available; its timestamp opens that point in the transcript.
3. Select **Save name**. If you can't tell who it was, close the editor without saving. To clear a saved name, expand the naming options and select **Remove name**.

The naming options also let you choose whether a name applies to the whole group or only selected transcript segments. Check that scope before saving, especially if a group contains more than one person's speech.

Names you choose carry into the AI Notes, search and Echo Agent's answers for that meeting. Echo won't replace a name you saved, or a voice you chose to leave unnamed, with a name of its own. A name typed here applies to this meeting only. Echo never guesses names from what people say.

## Play the recording

If **Record audio** was on, the Transcript tab has a player with skip back and forward and playback speed. Select **Delete audio** to remove the recording and keep the transcript.

## How long Echo keeps audio

Open **Settings › Account & workspace**:

| Setting | What it does |
| - | - |
| **Record audio** | On by default. The switch in meeting setup is the same setting |
| **After transcription** | **Keep audio** keeps recordings for playback. **Delete audio** removes new recordings once the transcript is saved |
| **Delete all audio** | Removes every recording you have. Transcripts and notes stay |

In Echo's cloud, recordings are encrypted in transit and at rest. A temporary, unencrypted copy on your computer is removed automatically within about 7 days.

## When a transcript is missing

If a meeting ended without a transcript but has a recording, select **Create transcript from recording**. If the transcript is incomplete, **Repair transcript** rebuilds it from the full recording and regenerates the notes.


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