> ## Documentation Index
> Fetch the complete documentation index at: https://smallestai-ff1e543d.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Code Examples

> Complete code samples for transcribing pre-recorded audio with Pulse STT

Below is a complete Python example demonstrating audio preprocessing, transcription with age/gender detection, emotion detection, and sentence-level timestamps (utterances).

<div style={{ maxHeight: '500px', overflowY: 'auto' }}>
  ```python theme={null}
  import os
  from pydub import AudioSegment
  from smallestai.waves import WavesClient

  client = WavesClient(api_key=os.getenv("SMALLEST_API_KEY"))

  def preprocess_audio(input_path, output_path):
      """
      Preprocess audio file to optimal format for Pulse STT:
      - Convert to 16 kHz mono WAV
      - Normalize audio levels
      - Remove leading/trailing silence
      """
      audio = AudioSegment.from_file(input_path)
      audio = audio.set_frame_rate(16000).set_channels(1)
      audio = audio.normalize()
      audio = audio.strip_silence(silence_len=100, silence_thresh=-40)
      audio.export(output_path, format="wav")
      print(f"Preprocessed audio saved to: {output_path}")
      return output_path

  def transcribe_with_features(audio_path):
      """
      Transcribe audio with age detection, emotion detection, and utterances.
      """
      response = client.transcribe(
          file_path=audio_path,
          model="pulse",
          language="en",
          word_timestamps=True,
          age_detection=True,
          gender_detection=True,
          emotion_detection=True,
          diarize=True
      )
      
      return response

  def process_results(response):
      """
      Extract and display transcription results.
      """
      print("=" * 60)
      print("TRANSCRIPTION RESULTS")
      print("=" * 60)
      
      print(f"\nTranscription: {response.get('transcription', 'N/A')}")
      
      if 'age' in response:
          print(f"\nAge: {response['age']}")
      if 'gender' in response:
          print(f"Gender: {response['gender']}")
      
      if 'emotions' in response:
          print("\nEmotion Scores:")
          emotions = response['emotions']
          for emotion, score in emotions.items():
              print(f"  {emotion.capitalize()}: {score:.2f}")
      
      if 'utterances' in response:
          print("\nUtterances (Sentence-level timestamps):")
          for i, utterance in enumerate(response['utterances'], 1):
              speaker = utterance.get('speaker', 'unknown')
              start = utterance.get('start', 0)
              end = utterance.get('end', 0)
              text = utterance.get('text', '')
              print(f"\n  [{i}] Speaker: {speaker}")
              print(f"      Time: {start:.2f}s - {end:.2f}s")
              print(f"      Text: {text}")
      
      if 'words' in response:
          print(f"\nWord-level timestamps: {len(response['words'])} words")

  if __name__ == "__main__":
      input_audio = "input_audio.mp3"
      preprocessed_audio = "preprocessed_audio.wav"
      
      try:
          print("Preprocessing audio...")
          preprocess_audio(input_audio, preprocessed_audio)
          
          print("\nTranscribing audio with age, emotion, and utterance detection...")
          result = transcribe_with_features(preprocessed_audio)
          
          process_results(result)
          
          if os.path.exists(preprocessed_audio):
              os.remove(preprocessed_audio)
              print("\nCleaned up temporary preprocessed file.")
              
      except FileNotFoundError:
          print(f"Error: Audio file '{input_audio}' not found.")
      except Exception as e:
          print(f"Error: {str(e)}")
  ```
</div>

## Prerequisites

Install required dependencies:

```bash theme={null}
pip install smallestai pydub
```

## Key Features Demonstrated

1. **Audio Preprocessing**: Converts audio to 16 kHz mono WAV, normalizes levels, and removes silence
2. **Age & Gender Detection**: Enables demographic analysis
3. **Emotion Detection**: Captures emotional tone with confidence scores
4. **Utterances**: Retrieves sentence-level timestamps with speaker labels
5. **Diarization**: Separates speakers for multi-speaker audio

## Expected Output

The script will output:

* Full transcription text
* Age and gender predictions
* Emotion scores (happiness, sadness, disgust, fear, anger)
* Sentence-level utterances with timestamps and speaker IDs