Skip to main content

SSML Support

Speech Synthesis Markup Language (SSML) support in TTS Wrapper is engine-dependent. Each engine has its own SSML implementation that maps to what the underlying service supports.

Using SSML​

Each TTS engine provides an SSML handler through the ssml property:

from tts_wrapper import PollyClient, PollyTTS

client = PollyClient(credentials=('region', 'key_id', 'access_key'))
tts = PollyTTS(client)

# Create SSML using the handler
ssml_text = tts.ssml.add('Hello <break time="500ms"/> world!')
tts.speak(ssml_text)

Engine-Specific SSML​

AWS Polly​

Supports the full range of SSML tags as documented in the Amazon Polly SSML Reference.

Google Cloud​

Follows the Google Cloud Text-to-Speech SSML Reference.

Microsoft Azure​

Implements SSML according to the Microsoft Speech Service SSML Reference.

IBM Watson​

Uses SSML as specified in the IBM Watson SSML Reference.

AVSynth (macOS)​

Converts SSML to AVSpeechSynthesizer commands:

  • <break> → [[slnc ms]]
  • <prosody> → Rate/pitch/volume commands
  • Other tags are stripped to plain text

eSpeak​

Supports basic SSML tags with its own extensions. See the eSpeak Documentation.

SAPI (Windows)​

Limited SSML support through the Windows Speech API.

ElevenLabs, Play.HT, Wit.ai​

These engines do not support SSML natively. SSML tags will be stripped and the plain text content will be used.

SSML Helper Methods​

Each engine's SSML handler provides helper methods for common operations:

# Clear any existing SSML content
tts.ssml.clear_ssml()

# Add text (may be plain text or SSML depending on engine)
tts.ssml.add("Text to speak")

# Get plain text (strips SSML tags)
plain_text = tts.ssml.get_text()

Best Practices​

  1. Check Engine Support: Always check the specific engine's documentation for supported SSML features
  2. Graceful Degradation: Provide plain text alternatives for unsupported SSML features
  3. Engine-Specific Features: Use engine-specific SSML features when needed for better control
  4. Test Thoroughly: Test SSML across different engines if your application needs to support multiple engines

Example: Engine-Specific SSML​

Here's how to handle SSML across different engines:

def speak_with_pause(tts, text: str, pause_ms: int = 500) -> None:
    """Demonstrate SSML handling across engines."""
    
    # Get the SSML handler for this engine
    ssml = tts.ssml
    ssml.clear_ssml()
    
    if isinstance(tts, PollyTTS):
        # Use Polly-specific SSML
        ssml_text = f'<speak>First part <break time="{pause_ms}ms"/> Second part</speak>'
    elif isinstance(tts, AVSynthTTS):
        # Use AVSynth command format
        ssml_text = f'First part [[slnc {pause_ms}]] Second part'
    else:
        # For engines without SSML support, just use plain text
        ssml_text = f'First part. Second part'
    
    tts.speak(ssml.add(ssml_text))

Next Steps​