TEST_INSTRUCTIONS.md

# Testing the TTS API

## How to Run Tests

### 1. Start the API server
```bash
python api.py
```

The API will start on `http://localhost:7860`

### 2. In another terminal, run the test suite
```bash
python test_api.py
```

## What the Tests Check

The test suite validates:

1. **Health Check** - Verifies the API is running and returns model information
2. **Speakers Endpoint** - Confirms available speakers are listed
3. **Input Validation** - Tests error handling for:
   - Empty text
   - Invalid speakers
4. **Audio Generation** - Tests TTS generation with:
   - Different speakers (Divya, Rani, Generic Female, etc.)
   - Urdu text input
   - Audio quality validation (format, duration, size)
5. **Audio Output** - Saves generated audio files to `test_outputs/` for manual inspection

## Test Output

Each test run generates:
- Console output with test results
- WAV audio files in `test_outputs/` directory for listening/verification

Example output:
```
============================================================
TTS API Test Suite
============================================================

1. Testing health check...
✓ Health check passed
  Available speakers: ['Divya', 'Rani', 'Rohit', 'Aman', 'Generic Female', 'Generic Male']
  Sample rate: 24000 Hz

2. Testing speakers endpoint...
✓ Speakers endpoint passed
  Speakers: ['Divya', 'Rani', 'Rohit', 'Aman', 'Generic Female', 'Generic Male']

3. Testing input validation...
✓ Empty text validation passed
✓ Invalid speaker validation passed

4. Testing TTS generation with different speakers...
✓ TTS generation successful
  Text: سلام دنیا
  Speaker: Divya
  Audio shape: (123456,)
  Sample rate: 24000 Hz
  Duration: 5.14 seconds
  Audio size: 246.91 KB
  Saved to: test_outputs/test_Divya_12_chars.wav
```

## Testing with CUDA

If you want to verify CUDA is being used:
1. Check the API startup logs - it should show:
   ```
   Using device: cuda
   GPU: NVIDIA GeForce RTX 4090 (or your GPU name)
   ```

2. Monitor GPU usage while tests run:
   ```bash
   nvidia-smi
   ```

## Manual Testing

You can also test manually with curl:

```bash
curl -X POST "http://localhost:7860/tts" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "سلام دنیا",
    "speaker": "Divya",
    "pitch": "Moderate",
    "rate": "Moderate"
  }' \
  --output test_audio.wav
```

Then play the audio file to verify quality.