We have been working very hard to improve the open-weights End of Turn detection model. Our ML team has made substantial improvements. One big problem we have seen is when agents need to collect specific data like phone numbers, email addresses, etc, they often stumble and add a lot of friction to the human-to-machine interaction experience. Also, you no longer need to pick an English or a multilingual model. You should use multilingual from now on. Check out our visualization tool that will let you compare the open weights models side by side: https://huggingface.co/spaces/livekit/eot-visualization?ref=blog.livekit.io I can hardly keep from spilling the beans on some other huge improvements we have in the lab right now but realtime AI is getting much better. Updated Model Blog Post with demo: https://blog.livekit.io/improved-end-of-turn-model-cuts-voice-ai-interruptions-39/ Plugin Code: https://github.com/livekit/agents/tree/main/livekit-plugins/livekit-plugins-turn-detector?ref=blog.livekit.io

Hey guys, I recently built and launched a code agent with Livekit, I am having problems with STT as it gets the wording wrong sometimes maybe because of it's in Italian. I am using gpt-4o-mini-transcribe as I found it the best so far for Italian. Is there any other model that's better than the one I am using for Italian? Thank you!

Has anyone successfully integrated LiveKit with Asterisk?

Hey guys , I hope you all are well . I am not sure if I am allowed to ask this question. But I am struggling with perhaps easy stuff for you . I tried so much but here my barrier is kicking in. I am not getting call ended events like transcripts or similar sensed to my webhook. Or precised: I am getting a message via webhook that a call has ended but always with missing transcripts. (I am using cartesia , deepgram etc.) Has anyone's idea why ? I can also share the codes if it's helpful . And another thing is , I don't get Vonage as phone provider properly working . (I am not using Twillio ) Has any one some secrets you can share with me ? Wish all happy Christmas Best Oli

LiveKit Blog Post xAI Blog Post We're excited to announce our partnership with xAI to bring Grok's voice technology to developers worldwide. Millions of people already talk to Grok through xAI's apps and in Tesla vehicles. Now you can build with the same technology through LiveKit.What makes Grok different? Most voice AI systems convert speech to text, process it, then convert back to speech. Grok uses true voice-to-voice processing—it understands and generates speech directly. This means lower latency and preserved emotional context like laughter, whispers, and natural conversational flow. The new LiveKit Agents plugin for Grok Voice Agent supports: - 5 distinct voice options - 100+ languages - Multimodal capabilities (generate images mid-conversation) - Function callingReal-world applications: This matters for customer service agents that need to detect tone and respond with empathy, healthcare applications where emotional context is critical, education platforms that adapt to student engagement, and accessibility services that require natural conversation. Get started in minutes: We've made integration simple. With just 4 lines of code and one command to install, you can add Grok's voice capabilities to your application. The LiveKit Agents plugin handles all the WebRTC transport, turn detection, and voice streaming. Try the playground at https://grok.livekit.io or check our docs at https://lnkd.in/gdw9EkSx