# Importing recordings

Create a session from an existing audio or video file. TalkBar transcribes it on your computer, separates speakers and writes the summary.

Canonical URL: https://docs.talkbar.app/sessions/importing-recordings


Already have a recording of a meeting, lecture or call? Import it, and TalkBar turns it into a session just like one you recorded live: a transcript by speaker, a timeline, a summary and a title. Transcription runs on your computer.

## Supported files

| | |
| --- | --- |
| **Formats** | mp3, wav, m4a, mp4, aac, flac, ogg, opus, mkv, webm |
| **Maximum size** | 1 GB |
| **Maximum length** | 6 hours |

Video files (mp4, mkv, webm) work too: TalkBar uses their audio track.

## Import a file

1. Click **New session** and set up the session as usual: choose a [template](/sessions/templates-and-skills), and optionally a project, series, context, summary language, length and privacy mode. See [Starting a session](/sessions/starting-a-session).
2. In the **Template** part, click **Upload audio file**.
3. Drop your file on the window, or click to browse for it.
4. Choose the options below, then click **Start import**.




| Option | What it does |
| --- | --- |
| **Speakers** | **Just you** is fastest: it skips speaker detection. **You + 1**, **Small (2–4)** and **Large (5–8)** separate who said what. |
| **Transcription model** | The speech model that transcribes the file. If the model isn't installed yet, you can download it from this window. |
| **Summary model** | The AI model used for the timeline entries and the summary. **Default** uses your configured model. For a Private AI session, the summary runs on a built-in local model. |

The session uses the rest of the setup from the New session window, such as the template, project and privacy mode.


If the speaker detection model isn't downloaded, the window tells you and offers to download it. See [Speakers](/voice/speakers).


## What happens during an import

The **Importing audio file** window shows each stage and how long it has taken:

1. **Decoding audio** and **Converting audio**: reads the file and saves it as the session's audio.
2. **Detecting speech**: finds the parts where people talk.
3. **Transcribing**: transcribes each part, and detects speakers.
4. **Saving transcript**.
5. **Summarizing timeline**: writes the timeline notes.
6. **Writing the summary**.
7. **Naming the session**: suggests a title.

You can keep the window open. When the import finishes, click **View session** to open it. Imported sessions are completed sessions, and their header shows **From audio file**.

Click **Process in background** to keep working while it runs. The session shows **Importing** in the Sessions list, and a notification says **Audio import complete** when it's done. Click **View session** in the notification to open it.

## Cancel an import

Click **Cancel import**. If some of the file was already transcribed, TalkBar keeps that part as a session. If nothing was transcribed yet, the empty session is removed.

## Things to know

- Only one import runs at a time. While it runs, you can't start a live session; you can still schedule one for later.
- The imported audio is kept with the session, so you can play it and use **Enhance transcript** later. See [Session details](/sessions/session-details#enhance-transcript).
- On the Free plan, an import counts toward the workspace's session limit. See [Plan limits](/sessions/overview#plan-limits).
- If the import fails, the window shows **Audio import failed** with the reason.

## Related

- [Starting a session](/sessions/starting-a-session)
- [Session details](/sessions/session-details)
- [Recording and transcription](/voice/recording-and-transcription)