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

# Voice, address, memory

> How to change Loli's voice, how she addresses you, what she remembers and how to clear the history

> [!NOTE]
>
> This page covers settings of the assistant arriving with the next update. Turning it on and commands: [Voice assistant](https://docs.getsoundcloud.app/en/use/voice.md). Needs [STAR](https://docs.getsoundcloud.app/en/star.md).

Everything is in **Settings -> Voice**.

## Voice

The **Reply voice** card: press a tile and that voice answers from now on. **Listen: name** plays a sample.

| Voice | Language |
| - | - |
| **Loli** | Russian |
| **Moriarty** | Russian |
| **Heart**, **Michael** | English |

| Setting | What it does |
| - | - |
| **Voice from our servers** | Only for Loli and Moriarty. For a weak computer: the voice model is not downloaded, a reply may wait in a queue |
| **Tone and pace** | **Speed** and **Pitch** sliders |
| **Speak on** | **With music**, **Device** (pick speakers or headphones), **Where I listen** |
| **Music under voice** | **Music level** while the assistant speaks |
| **Reply by voice** | Turn off to get replies on screen only |

## Address and role

The **You and the assistant** card. It works while **Talk with the character** is on; when it is off there is a **Turn on conversation** button.

| Setting | Options |
| - | - |
| **Gender** | **he**, **she** |
| **Formality** | **casual**, **formal**. Not chosen: it follows the role (**Comes with the role**) |
| **Name or nickname** | Any word |
| **Role** | **Friend**, **Assistant**, **Butler**, **Mentor**, **Custom** |

**Custom**: type it into **Who should I be to you?**. Pressing the chosen one again clears the choice. **Say hi** checks how it sounds.

## Conversation and memory

The **Conversation** card.

* **Talk with the character** (on by default). Phrases that are not commands go to the conversation. Our server and the model provider get the text of the phrase, what is playing and queued, and the facts she remembers. Audio never leaves the computer. Turn it off and only commands remain.
* **Memory**: up to 30 facts about you. Say "remember that ..." or press **Add a fact**. **Forget** erases one fact, **Forget all** erases everything after you confirm with **Yes, forget all**.

More: [privacy policy](https://getsoundcloud.app/en/privacy).

## History

The **History** card: what the assistant heard and what it did or answered. It shows the latest rows, the rest are under **N more**; **Show less** folds them.

* **Undo** reverts what was done, while that is possible.
* The flag **Understood wrong** marks a row.
* **Clear** erases the whole history.

## Models

The **Models** card shows what the assistant needs: **Speech detector**, **Recognition**, **Speech synthesis**, **Wake word**, **Pronunciation**, **Smart parsing**, **Speech engine**. Each has a size, optional ones are marked **optional**. Space used: **N on disk**.

| Button | When |
| - | - |
| Download icon, tooltip **Download name** | The model is missing. **Download · size** at the top gets everything missing |
| **Resume** | The download was interrupted |
| **Download again** | **Model damaged** |
| Trash icon on hover, **Delete name** | Free up space. Without a needed model the assistant will not work until you download it again |

Models download from our storage the first time you turn the assistant on. **Licenses** at the bottom shows where data and models come from.


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