# Settings Guide

This guide explains all the settings available in the Live Transcription App.

## Accessing Settings

Click the **⚙️ Settings** button in the top-right corner of the main window to open the settings page.

## Settings Overview

### API Configuration

#### OpenAI API Key
- Required for using Whisper Remote transcription
- Get your key from: https://platform.openai.com/api-keys
- Stored securely and never shown in plain text after saving

#### ElevenLabs API Key
- Required for using ElevenLabs Scribe transcription
- Get your key from: https://elevenlabs.io/api
- Stored securely and never shown in plain text after saving

### Audio Processing

#### Whisper Model
Select the Whisper model size based on your needs:
- **Tiny**: Fastest, least accurate (39M parameters)
- **Base**: Good balance (74M parameters) - Default
- **Small**: More accurate (244M parameters)
- **Medium**: High accuracy (769M parameters)
- **Large**: Highest accuracy, slowest (1550M parameters)

#### GPU Acceleration
Choose your GPU acceleration method:
- **Disabled (CPU only)**: Uses CPU for processing
- **NVIDIA CUDA**: For NVIDIA GPUs (RTX, GTX, etc.)
- **AMD ROCm**: For AMD GPUs (Linux only)
- **Intel OpenVINO**: For Intel Arc and integrated GPUs

Use the **Detect GPU** button to automatically detect and select the best option.

#### Audio Threshold
- Controls sensitivity to sound (0.001 to 0.1)
- Lower values = more sensitive (picks up quiet sounds)
- Higher values = less sensitive (only loud sounds)
- Default: 0.01

#### Transcription Speed
Controls how often audio chunks are processed:
- **Very Fast (1s)**: Near real-time, may cut off words
- **Fast (2s)**: Good responsiveness - Default
- **Balanced (3s)**: Better for complete phrases
- **Quality (5s)**: Best accuracy for long sentences

#### Ambient Noise Calibration
1. Click **Start Calibration**
2. Stay quiet for 10 seconds
3. The app measures your room's background noise
4. Automatically sets optimal threshold
5. Shows current ambient level when calibrated

### Display Settings

#### Default Backend
Choose which transcription service to use by default:
- Whisper (Local)
- Whisper (Remote)
- ElevenLabs (Scribe)

#### Text Duration
How long transcribed text stays visible (1-60 seconds)

#### Text Color
Color of the transcribed text

#### Background Color
Background color of the transcription area

#### Font Family
Choose from system fonts or Google Fonts

#### Font Size
Adjust text size from 16px to 72px

### Server Settings

#### Server Port
- Default: 3000
- Change if port conflicts with other applications
- Requires app restart after changing

## Settings Storage

Settings are stored in:
- **Windows**: `%USERPROFILE%\.live-transcription\settings.json`
- **macOS**: `~/.live-transcription/settings.json`
- **Linux**: `~/.live-transcription/settings.json`

## Testing Features

### Test API Connections
- Tests OpenAI API key validity
- Tests ElevenLabs API key validity
- Shows success/failure status

### Detect GPU
- Automatically detects available GPUs
- Shows NVIDIA CUDA availability
- Shows AMD ROCm availability
- Shows Intel GPU availability
- Recommends best option

## Best Practices

1. **For Best Performance**:
   - Use GPU acceleration if available
   - Choose appropriate model size for your hardware
   - Use "Fast (2s)" transcription speed

2. **For Best Accuracy**:
   - Use larger Whisper models (Small/Medium/Large)
   - Use "Quality (5s)" transcription speed
   - Calibrate ambient noise in your environment

3. **For Low-End Hardware**:
   - Use Tiny or Base models
   - Disable GPU if causing issues
   - Increase audio threshold to reduce processing

## Troubleshooting

### Settings Not Saving
- Check write permissions for settings directory
- Try "Reset to Defaults" if settings are corrupted

### GPU Not Detected
- Ensure GPU drivers are installed
- For NVIDIA: Install CUDA toolkit
- For AMD: Install ROCm (Linux only)
- For Intel: Install OpenVINO runtime

### High CPU Usage
- Enable GPU acceleration if available
- Use smaller Whisper model
- Increase audio threshold
- Use longer transcription chunks

### API Keys Not Working
- Use "Test API Connections" to verify
- Check for typos or extra spaces
- Ensure API keys have correct permissions
- Check API usage limits