Eleven v4 TurboNow in ElevenAgents

Eleven v4 is tuned for produced content like audiobooks and voiceovers. Eleven v4 Turbo is tuned for agents and real-time use with a median inference latency of ~100 ms. Both share the same architecture, emotional range, and support for Professional Voice Clones.

Eleven v4 Turbo is our fastest model yet. Median inference latency is ~100 ms and median time to first speech is ~150 ms.

ElevenAgents picks up on how a caller is feeling. With Eleven v4 Turbo, it can adjust its delivery to match, with elevated expression and consistency on long calls.

Eleven v4 Turbo supports 90+ languages and accents. An agent can switch languages mid-call when a caller does, keeping the same voice throughout.

Yes. Both Instant and Professional Voice Clones work with Eleven v4 Turbo, and a clone performs with the same expressive range as any library voice. Every clone requires verified consent from the voice's owner.

Yes. It supports telephony output and streams audio in the format your phone system needs. Agents in ElevenAgents can take inbound and outbound calls over standard phone lines.

The agent stops mid-word, takes in what the caller said, and answers the new question in the same turn. There's no restart and no repeating the interrupted line.

Yes. Eleven v4 Turbo is available through the API with bidirectional streaming, for teams building agents on their own stack.