Eleven Flash v2.5
ElevenLabs' low-latency TTS model built for real-time conversational agents
Eleven Flash v2.5 is ElevenLabs’ fastest text-to-speech model, built for voice agents and other applications where response time matters more than raw audio polish. It generates speech in about 75 milliseconds, not counting network and application overhead, and covers 32 languages including additions like Hungarian, Norwegian, and Vietnamese that weren’t in the earlier Flash v2. ElevenLabs prices it at roughly half the cost per character of its flagship Multilingual v2 model, which makes it the default choice for conversational AI products, live customer support bots, and other real-time voice interfaces where a 300-millisecond model would feel sluggish. The tradeoff is expressiveness: Flash v2.5 sounds clean and fast but doesn’t carry the emotional range ElevenLabs built into Eleven v3.