# Voice Models & Memory Management

Aura uses offline speech recognition models (powered by Vosk). Models are available for over 20 languages and come in two distinct tiers to fit any hardware.

---

## 1. Model Tiers

| Tier | Folder Pattern | RAM Usage | Best For |
|---|---|---|---|
| **Small Models** | `vosk-model-small-*` | ~100–250 MB | Systems with < 8 GB RAM, Laptops, Raspberry Pi |
| **Standard Models** | `vosk-model-*` | ~1.5–4.5 GB | Systems with ≥ 8 GB RAM, Desktop workstations |

---

## 2. Supported Languages

Vosk provides official offline models for English, German, French, Spanish, Italian, Portuguese, Russian, Chinese, Japanese, and many others.

- **Browse & Download Models:** [Official Vosk Model Directory](https://alphacephei.com/vosk/models)

---

## 3. How to Install and Switch Models

### Step 1: Download and Extract
Download the model archive for your language and extract it into the `models/` directory:

```bash
modelos de CD
wget -q --show-progress https://alphacephei.com/vosk/models/vosk-model-small-de-0.15.zip
descompacte -q vosk-model-small-de-0.15.zip
rm vosk-model-small-de-0.15.zip
cd ..

Step 2: Set Active Model

Write the exact folder name into config/model_name.txt:

echo "vosk-model-small-de-0.15" > config/model_name.txt

Step 3: Set in config/settings.py

open config/settings.py change model there

PRELOAD_MODELS = ["modelo vosk .. "]
__CODE_BLOCO_3__
sudo swapoff -a
sudo fallocate -l 8G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
echo '/swapfile nenhum swap sw 0 0' | sudo tee -a /etc/fstab

Restart Aura to apply the new model.

4. Troubleshooting & Virtual RAM (Swap)

If Aura’s memory watchdog triggers or the application runs out of memory, follow these steps:

Option B: Configure Virtual RAM (Swap / Paging File)

Linux (Create an 8 GB Swapfile):

sysctl vm.swapusage
__CODE_BLOCO_5__