> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aurinfer.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Voice Engine
> Configure your character's voice — model, audio quality settings, speaking frequency, expression instructions, and voice call behavior.
# Voice Engine
This is where you configure **how your character sounds** — the actual voice it uses, how expressive or stable that voice is, how often it speaks vs. types, and how it behaves in live voice calls.
**How to get here:**
> Go to your character's dashboard → left sidebar → click **Universal** → click **Voice Engine**
The full path is: **Universal → Voice Engine**
This section has **five sub-pages** shown at the top of the Voice Engine area:
1. **Overview** — details about the current voice
2. **Settings** — sliders to adjust how the voice sounds
3. **Frequency** — how often the character uses voice vs. text
4. **Instructions** — how to use audio tags for emotion/expression
5. **Preferences** — how the character behaves in voice channels
At the very top of every Voice Engine page, you'll see your **currently assigned voice** displayed with its provider badge (for example, **Eleven v3** from ElevenLabs) and a toggle to enable or disable it.
***
## Overview
**Full path: Universal → Voice Engine → Overview**
This is the information page about your current voice — think of it like the voice's "profile page."
### Voice Details
| Field | What it means |
| ------------ | -------------------------------------------------------------------------------------------------------------------- |
| **Voice ID** | A unique code that identifies this specific voice. You can copy it by clicking it. |
| **Model** | The text-to-speech (TTS) model being used to generate audio. Example: `eleven_v3`. |
| **Language** | The main language this voice was trained in. Example: `En` for English. |
| **Accent** | The accent of the voice (if it has one). Example: `persian`. You can remove the accent by clicking the ✕ next to it. |
| **Gender** | The voice's gender classification. Example: Male. |
### Voice Latency
These numbers tell you how fast or slow the voice generates audio:
| Metric | What it means |
| ----------- | ---------------------------------------------------------------------------------- |
| **Average** | How many seconds it typically takes to generate a voice response. Lower is better. |
| **P95** | The "worst case" time — 95% of all responses come in faster than this number. |
These numbers only measure how long it takes to create the audio itself. The total time before your character speaks also includes time for the AI to write what to say — that's separate.
### Training Samples
At the bottom of the Overview page, you'll see a list of audio files (MPEG format). These are the original recordings the voice was cloned from. Each file shows its length and file size.
### Changing Your Voice
Click the **Edit Voice** button on this page to switch to a different voice for your character.
***
## Settings
**Full path: Universal → Voice Engine → Settings**
These are **sliders** that let you fine-tune how the voice sounds. Moving them left or right changes the character's audio quality:
| Setting | What it does | Simple explanation |
| -------------------- | ------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------- |
| **Stability** | How consistent the voice sounds from sentence to sentence. | High = the voice sounds very even and uniform. Low = the voice has more natural variation, like a real person. |
| **Similarity Boost** | How closely the output matches the original voice recording it was cloned from. | High = sounds more like the original clone recording. Low = sounds slightly different. |
| **Style** | How much the voice exaggerates its emotional style and emphasis. | High = dramatic, expressive. Low = flat and neutral. |
| **Speed** | How fast the character talks. | Low = slow and deliberate. High = fast. |
| **Pitch** | How high or low the voice sounds overall. | Move up for a higher pitch, down for a deeper voice. |
**For natural conversation:** Try Stability at 0.6, Style at 0.4. This gives slight variation without sounding unstable.
**For dramatic characters:** Lower Stability (0.3) and higher Style (0.6–0.8) makes the voice more expressive and theatrical.
***
## Frequency
**Full path: Universal → Voice Engine → Frequency**
This controls **how often your character responds with a voice message** instead of a regular text message (in Discord text channels where voice is enabled).
There are two modes. You can only use one at a time:
### Let Character Decide (toggle)
When this toggle is **ON**, the character will decide on its own whether to send a voice message or a text reply based on the conversation context. The manual slider below becomes disabled.
### Manual Frequency Control (slider)
When "Let Character Decide" is **OFF**, you drag a slider to set the exact rate:
| Slider position | What happens |
| ------------------------- | ---------------------------------------------------------------- |
| All the way **left** (0) | The character **never** sends voice messages — always text only. |
| In the **middle** (\~0.5) | The character sends voice messages about **half the time**. |
| All the way **right** (1) | The character **always** responds with voice — never text. |
***
## Instructions
**Full path: Universal → Voice Engine → Instructions**
This is a text box (up to 1000 characters) where you write instructions for how the voice uses **audio tags** to show emotion.
### What are Audio Tags?
Audio tags are special words you put in square brackets `[like this]` that tell the voice model to add a specific sound or change how it speaks. The AI includes these tags in the text before it's turned into audio.
**Examples of audio tags:**
| Tag | What it does |
| ------------------ | -------------------------------------------------- |
| `[whisper]` | Makes the voice quiet and hushed — like whispering |
| `[yells]` | Makes the voice louder and more intense |
| `[laughs]` | The voice laughs |
| `[sigh]` | The voice sighs |
| `[crying]` | The voice sounds like it's crying |
| `[breathes]` | Adds a breath sound |
| `[applause]` | Adds an applause sound effect |
| `[British accent]` | Changes the accent mid-speech |
| `[echoing]` | Adds an echo effect |
| `[explosion]` | Adds an explosion sound effect |
### How to write the Instructions
In the text box, write rules that tell the AI when to use these tags. For example:
```
Use [laughs] before sarcastic remarks.
Insert [breathes] occasionally for a dramatic feel.
Use [yells] when responding to rude or aggressive messages.
Add [sigh] before answering boring or repetitive questions.
```
The character will follow these rules every time it generates a voice response.
***
## Preferences
**Full path: Universal → Voice Engine → Preferences**
These are two simple on/off toggles that control what your character does in Discord voice channels:
| Setting | What it does |
| ----------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Auto-join Voice Channels** | When ON, the character automatically joins a voice channel as soon as real members are in it — without being invited. When OFF, someone has to manually ask it to join. |
| **Auto-leave Voice Channels** | When ON, the character automatically leaves the voice channel when all human members have left — so it's not sitting alone in an empty channel. When OFF, it stays until someone asks it to leave. |