Just wrapped up a fun holiday project - a voice agent where you can actually call Santa Claus! What I built: A character-driven voice agent that maintains Santa's warm, jolly persona throughout real phone conversations. Adults can call in and reconnect with a bit of Christmas magic. Tech stack: - LiveKit - Real-time voice communication + SIP support for telephony - Speechmatics - STT with custom vocabulary for Christmas terms ("Nice List", "Naughty List", etc.) - ElevenLabs - Custom Santa voice using multilingual v2 - OpenAI - Conversation generation while maintaining character consistency - Twilio - Optional PSTN integration for real phone calls Demo: https://www.youtube.com/shorts/SWnhkz7v6AQ Code is open source if anyone wants to build their own holiday hotline: https://github.com/speechmatics/speechmatics-academy/tree/main/use-cases/05-santa-voice-agent Happy holidays everyone!

I saw that on Twitter and it made me laugh 😂 thanks for sharing.

So I have been working on www.voxform.xyz, it basically lets you create forms that acts as voice interviews to collect responses. It uses a livekit cloud agent, deepgram from stt, openai models for llm and tts. Now for Christmas I designed a voice akin to Santa. So now you can share your wished directly to him. Here is the link to the demo voice form - https://www.voxform.xyz/s/1C3ETZsGbM8M

Very cool I had json being outputted at one point for wish list but decided to keep it simpler. This looks good too!

Do gve it a try. I already have the agent ready to accept answers to the form questions. So it primarily about getting voice right. And i think your Santa voice is better!

@dazzling-soccer-1833 I love this overall pretty good! I agree voice could be improved! Love the idea of form might readopt this into Academy if you good with that!

Thank you Edgar! Trying to reduce latency and add voices, plus improve them. But what do you mean by academy?

Not sure what you using for stt but recommend checking out Speechmatics one we just made PR for livekit awaiting merge once that is in we should be as fast as the deepgram! Also have different ways to handle EOU. Check out some samples here: https://github.com/speechmatics/speechmatics-academy/tree/main/basics/08-voice-agent-turn-detection

I have been using Deepgram stt and livekit EOU. Will surely experiment with Speechmatics ✌️