Kokoro-ONNX TTS Demo

Optimized ONNX implementation with Voice Blending

Understanding Text Splitting

The splitting pattern controls how Kokoro breaks your text into manageable chunks for processing.

Common patterns:

  • \n+: Split on one or more newlines (paragraphs)
  • (?<=[.!?])\s+: Split after periods, question marks, and exclamation points (sentences)
  • [,;]\s+: Split after commas and semicolons
  • $^: Special pattern that won't match anything (processes the entire text as one chunk)

Benefits of splitting:

  • Better phrasing and natural pauses
  • Improved handling of longer texts
  • More consistent pronunciation across chunks
Split Text Using

Select how to split your text into chunks
Language

Select the language for text processing
Primary Voice

Select primary voice for synthesis
Secondary Voice for Blending

Select secondary voice to blend with primary voice
0 1

Voice blending lets you combine characteristics of two voices.

  • A 50:50 blend gives equal weight to both voices
  • Higher values emphasize the primary voice
  • Lower values emphasize the secondary voice
0.5 1.5

Performance Testing

When enabled, clicking "Generate Audio" will run performance tests instead of generating audio. Tests compare different processing approaches to identify the most efficient method.

Use this to optimize your implementation based on your specific hardware and text content.