IStreamingTts<T>

Interfaces · AiDotNet.TextToSpeech.Interfaces

Interface for TTS models that support streaming/chunked synthesis with low latency.

Implements: ITtsModel<T>

How It Works

Streaming TTS models can begin outputting audio before the full utterance is processed, enabling low-latency applications like conversational AI:

  • CosyVoice 2: 150ms first-packet latency with streaming flow matching
  • Chatterbox: real-time streaming with emotion control
  • XTTS-v2: chunked streaming for voice cloning

Properties

Property Summary
bool HasMoreChunks { get; } Gets whether there are more audio chunks available.
int FirstPacketLatencyMs { get; } Gets the target first-packet latency in milliseconds.

Methods

Method Summary
Tensor<T> SynthesizeFirstChunk(string text, int chunkSize) Synthesizes audio in streaming chunks, returning the first available audio chunk.
Tensor<T> SynthesizeNextChunk() Gets the next audio chunk from an ongoing streaming synthesis.