ElevenLabs launches Conversational AI 2.0 with sub-300ms voice response latency

Read the original article →

What happened

ElevenLabs released Conversational AI 2.0 on July 8, cutting end-to-end voice response latency below 300 milliseconds for English and adding support for 32 languages. The platform combines speech recognition, language-model reasoning and text-to-speech in a single API with turn-taking detection.

Pricing starts at $0.08 per minute for hosted conversations. ElevenLabs said the release targets customer-service and voice-agent developers.

Why it matters

Sub-300ms latency is the threshold where voice agents feel conversational rather than sluggish. Bundling STT, LLM and TTS in one API reduces integration work for teams building voice products.

It also puts ElevenLabs into more direct competition with xAI's Grok speech APIs and established providers.

MintedBrain take

Latency claims need testing on your network and languages. Run a pilot with real user audio before committing a production voice agent to any single provider.

References

This article was originally published at ElevenLabs Docs. For the full piece, read the original article.

Discussion

  • Loading…

← Back to News