WebRTC Audio for AI Companion
I'm looking for a good platform to use to build a real-time (meaning streamed and interruptible) voice AI companion. Vapi.ai is one of those platforms but at $.05/minute, it's too expensive for a use case where the customer might interact for a hour or more a day.
I initially thought daily.co might be a good alternative. Your blog post "How to talk to an LLM" is helpful. But then I saw the blog post about your partnership with VAPI.ai, and how that's a great combination.
Is the difference between Daily.co and VAPI, that while either will be a good platform, more of the coding having to do with endpointing and barge-in is already built into the platform on Vapi.ai?
Thx
Best Answer
-
Vapi has lots of good low-level handling for things like endpointing and interruptions, and also has higher level tools and dashboards that make it easy for a broad range of developers to build, test, and scale.
If you have a use case that you don't think Vapi is the perfect fit for, maybe you want to take a look at the open source framework we just launched, designed for people prototyping new real-time, multimodal AI tools:
The framework includes example code for several simple voice agents, including how to stream data between the stages in the processing pipeline, how to implement fast responses with voice activity detection, and how to stop and reset a pipeline so you can handle use interruptions.
Here's a simple chatbot:
1
