Streaming TTS - RunAnywhere Documentation

synthesizeStream emits AudioChunk values as they are produced, so playback can start on the first chunk instead of waiting for the last.

let chunks = try await RunAnywhere.tts.synthesizeStream(longText)

for try await chunk in chunks {
    player.enqueue(chunk)
}

Signature

func synthesizeStream(_ text: String, options: TtsOptions? = nil) async throws
    -> AsyncThrowingStream<AudioChunk, Error>

When it is worth it

For a sentence, speak is simpler and the latency difference is not noticeable. For a paragraph or a model reply being streamed as it generates, streaming synthesis is what keeps the gap between “the model finished thinking” and “the user hears something” short.

Pairing it with a streaming completion

Synthesize each sentence as the LLM finishes it, rather than waiting for the whole reply:

var sentence = ""

for try await event in try await RunAnywhere.llm.generateStream(prompt: prompt) {
    guard case .token(let token) = event else { continue }
    sentence += token.text

    if token.text.contains(where: { ".!?".contains($0) }) {
        let handle = try await RunAnywhere.tts.speak(sentence)
        await handle.waitForPlayout()
        sentence = ""
    }
}

if !sentence.isEmpty {
    _ = try await RunAnywhere.tts.speak(sentence)
}

For a full conversation rather than one reply, use the voice session, which handles turn-taking and interruption for you.

Stopping

await RunAnywhere.tts.stop()