WebSocket TTS streaming for voice agents: LLM tokens in, speech out, measured in October 2026
Since 9 October 2026, SpeechifyAI takes an LLM's reply over a WebSocket as it is written. From Google Cloud us-east4 that day, the first word was audible 49 ms after the LLM finished its first sentence, against 68 ms with one HTTP request per sentence.
SpeechifyAI team