I am integrating openAI realtime API using node js and ARI. I set talk detect in dialplan but that is not working accurately I am receiving ChannelTalkingStarted event continuously if there is background noise.I am developing AI agent so can anyone please provide me solution for this.
The only builtin functionality is exactly that, and the configuration it allows. There is no further advanced support, something like that would need to be done either by extending func_talkdetect or putting your own logic in the ARI application.
Hello @jcolp thanks for the reply.
Okay so is there any other ways available in asterisk regarding VAD functionality or is there any way regarding I can send direct live audio stream to open AI realtime API in the format of g711 alaw or ulaw?
As I stated, there aren’t other options built in. I would also search for existing posts regarding ARI and sending media to it, and then forwarding it on. Your question is honestly being asked at minimum weekly at this point, and I’ve answered it so many times at this point.
okay @jcolp Thanks for reply. But in this AI era everyone wants to create virtual AI agents or voice bot so it is really good if asterisk provides integrations like this.
Asterisk is a toolkit, we provide the underlying fundamentals for things such as doing this. External media in ARI supports shipping the audio to your ARI application. What you do with it after that is up to you. Send it to OpenAI, use another service for speech to text and do text based, up to you. This works, people are doing it for voice bots/AI agents. If what you’re wanting is an out of the box experience that is a virtual AI agent/voice bot, Asterisk itself won’t give you that without work. We’re just a component to make such a thing.
@jcolp really appreciate for your help .
does External media formats send in the formats supported by open AI or do I have to convert it?
The documentation[1] has the options that are supported when creating an external media channel. One of those being format, to allow specifying ulaw or alaw for example.
Okay @jcolp Thanks. Let me check it out and give it a try.