Generate speech using Kokoro TTS models

Use This When

• Building voice assistants that need fast, natural-sounding text-to-speech responses
• Creating conversational UIs where low-latency TTS improves user experience
• Implementing audio feedback systems for accessibility or hands-free operation
• Generating voice narration for automated content creation or alerts

What It Does

• Converts text strings to audio using Kokoro neural TTS pipeline
• Supports multiple voice presets via configurable voice parameter (e.g., af_heart)
• Returns AudioFrame at 24kHz sample rate (tensor shape [samples, channels], typically [N, 1])
• Handles empty input gracefully by returning silent audio frame

Works Best With

• query-llm → this component → output-audio-file or audio playback for voice responses
• Chatbot systems → this component → real-time audio streaming to users
• Integration with detect-voice-activity for bidirectional voice conversations
• Alert systems that need spoken notifications rather than visual displays

Caveats

• Fixed 24kHz output sample rate; resampling required if downstream expects different rate
• Voice quality and availability depends on Kokoro model; verify voice parameter validity
• Real-time factor (RTF) varies by text length and hardware; GPU strongly recommended
• Language code parameter affects pronunciation; ensure alignment with input text language