Flash is ElevenLabs' latest model for generating human-like text-to-speech (TTS) in an impressively low latency of 75ms plus application and network delays, designed for use in conversational voice agents. It is available in two versions: Flash v2, which supports English, and Flash v2.5, which supports 32 languages. Both versions are accessible through the company's Conversational AI platform or directly via API and are priced at 1 credit for every 2 characters. Although Flash has slightly lower quality and emotional depth compared to the Turbo models, it excels in speed, consistently outperforming similar ultra-low-latency models in blind tests. ElevenLabs encourages users to explore their developer guides to optimize model use and looks forward to innovations in conversational interactions enabled by Flash's capabilities.