6s delay building an AI receptionist with 3CX Call Flow Designer (CFD) + n8n (VoiceInput latency)
When investigating ways of programming an AI receptionist I came across the program Call Flow Designer. I created a flow containing a VoiceInput node that would transcribe the voice of the caller, send it to n8n through an httpRequest node and receive the action that the receptionist should do, either to keep talking, transfer the call or hang up.
Between the moment the caller stopped talking and the AI processed the httpRequest, there were 6 seconds of silence. Through a series of loggers I realized the 6 seconds occurred right before the voiceInput node decided whether the input was valid or not. I tried to fill up that void with music but due to the way the node is programmed it is impossible.
Has anyone tried this and had better results using something I'm not aware of? I'm fairly new to the 3cx software so any help would be appreciated.
I also post this so that if anyone hits the same problem, they know it has happened before and hopefully there's an answer below. In my case I ended up giving up on the CFD for this and switched to a streaming solution to get a more natural flow of conversation.