I don't mind music-on-hold as much except when they break with a human voice at intervals saying inane things such as "your call is important to us. Thank you for holding" Which makes me think the call has been answered. If there's not actual call queue information, just play music so I can ignore it and continue working without that NOP interrupt.

My local dealership adopted one of these for their service department. Prior to the AI assistant, you would call and immediately be placed on hold because the same people serving customers at the desk were answering the phone. The AI picks up right away and can schedule an oil change. It's fantastic and I'd love to see more of this. Of course it also has the ability to escalate to a human for the things it doesn't have capabilities for.

For narrow use cases like this I personally don't mind these tools.

Good customer support lines? Is there a reason why it can't provide good support. I often use chatgpt's voice function.

This might be good for Back-office hands-free tool access, for employees who are on the road. (they shouldn't be looking at the screen, and they might be limited to voice calling due to coverage issues besides). Aka: really weird terminal.

> What is the application of this?

spammers, scammers and horrible customer support lines.

[dead]

[flagged]

Voice-controlled phone systems are hugely rage-inducing for me. I am often in loud setting with background chatter. Muting my audio and using a touchtone keypad is so much more accurate and easy than having to find a quiet place and worrying that somebody is going to say something that the voice response system detects.

One problem is once you’re in deep building a phone IVR workflow beyond X or Y (yes, these are intentional), callers don’t care about some deep and featured input menu. They just mash 0 or pick a random option and demand a human finish the job and transfer them - understandably.

When you’re committed to phone intent complexity (hell), the AI assisted options are sort of less bad since you don’t have to explain the menu to callers, they just make demands.

The problem here is that if it's something a voice assistant can solve, I can solve it from my account. I'm calling because I need to speak to an actual human.

Well, the future is here: https://www.youtube.com/watch?v=HbDnxzrbxn4

> Looked at a repo and it seems this project was heavily AI generated

That's the first thing that I noticed too.

It's gotten to the point that my body subconsciously rejects bullet lists and headings that start with emojis.

[dead]

So, there is no straight forward way that I can think of. But, what I would do is set a channel variable and log it somewhere:

exten => s,n,Set(VM_UNIQUEID=${UNIQUEID}) exten => s,n,VoiceMail(${EXTEN}@default)

If you are using AGI or ARI, you can log it somewhere useful so you can correlate later.

If you are using a more vanilla configuration I’d say use the voicemail metadata .txt file that will be in the same folder as the recording to get info to find the CDR. It has things like callerid, origmailbox, origdate (or maybe it’s origtime), and duration. origmailbox should match the CDR destination and the orig time should also match. Haven’t done this specifically. But, I’m hoping I’m pointing you in the right direction.

I work with Freeswitch almost exclusively these days. But, my first experience with VoIP was Asterisk and a huge Perl AGI file keeping everyone talking to each other. Those were good time!

I've generally observed latency of 500ms to 1s with modern LLM-based voice agents making real calls. That's good enough to have real conversations.

I attended VAPI Con earlier this year, and a lot of the discussion centered on how interruptions and turn detection are the next frontier in making voice agents smoother conversationalists. Knowing when to speak is a hard problem even for humans, but when you listen to a lot of voice agent calls, the friction point right now tends to be either interrupting too often or waiting too long to respond.

The major players are clearly working on this. Deepgram announced a new SOTA (Flux) for turn detection at the conference. Feels like an area where we'll see even more progress in the next year.

Sesame was the fastest model for a bit. Not sure what that team is doing anymore, they kind of went radio silent.

https://app.sesame.com/

Absolutely not.

500-1000ms is borderline acceptable.

Sub-300ms is closer to SOTA.

2000ms or more means people will hang up.

Been experimenting with having a local Home Assistant agent include a qwen 0.5B model to provide a quick response to indicate that the agent is "thinking" about the request. It seems to work ok for the use case, but it feels like it'd get really repetitive for a 2 way conversation. Another way to handle this would be to have the small model provide the first 3-5 words of a (non-commital) response and feed that in as part of the prompt to the larger model.

From my personal experience building a few AI IVR demos with Asterisk in early 2025, testing STT/TTS/inference products from a handful of different vendors, a reliable maximum latency of 2-3 seconds sounds like a definite improvement. Just a year ago I saw times from 3 to 8 seconds even on short inputs rendering short outputs. One half of this is of course over-committed resources. But clearly the executional performance of these models is improving.

Very randomly and personally I appear to have experimented with that few months ago, based on a Japanese advent calendar project[1] - the code is all over the place and only works with Japanese speeches, but the gist is as follows. Also in [2].

The trick is to NOT wait for the LLM to finish talking, but:

  1 at end of user VAD, call LLM, stream response into a buffer(simple enough)   
  2 chunk the response at [commas, periods, newlines], and queue sentence-oid texts  
  3 pipe queued sentence-oid fragments into a fast classical TTS and queue audio snippets   
  4 play queued sentence-oid-audio-snippets, maintaining correspondence of consumed audio and text 
  5 at user VAD, stop and clear everything everywhere, undoing queued unplayed voice, nuking unplayed text from chat log 
  6 at end of user VAD, feed the amended transcripts that are canonical to user's ears to step 1
  7 (make sure to parallelize it all)

  user: "what's the date today"
  sys:  "[today][is thursday], [decem"
  user: "sorry yesterday"
  sys:  "[...uh,][wednesday?][Usually?]"

  1: https://developers.cyberagent.co.jp/blog/archives/44592/
  2: https://gist.github.com/numpad0/18ae612675688eeccd3af5eabcfdf686

Microsoft Foundry's realtime voice API (which itself is wrapping AI models from the major players) has response times in the milliseconds.

No, there are models with sub-second latency for sure

Just try Gemini Live on your phone. That's state of the art

Perhaps you didnt read that these are "production-ready golden baselines validated for enterprise deployment."

How does their golden nature not dissuade these concerns for you?

[dead]

She seems to be doing well

https://m.youtube.com/watch?v=wairnc-2Hyo

One easy way to build voice agents and connect them to Twilio is the Pipecat open source framework. Pipecat supports a wide variety of network transports, including the Twilio MediaStream WebSocket protocol so you don't have to bounce through a SIP server. Here's a getting started doc.[1]

(If you do need SIP, this Asterisk project looks really great.)

Pipecat has 90 or so integrations with all the models/services people use for voice AI these days. NVIDIA, AWS, all the foundation labs, all the voice AI labs, most of the video AI labs, and lots of other people use/contribute to Pipecat. And there's lots of interesting stuff in the ecosystem, like the open source, open data, open training code Smart Turn audio turn detection model [2], and the Pipecat Flows state machine library [3].

[1] - https://docs.pipecat.ai/guides/telephony/twilio-websockets [2] - https://github.com/pipecat-ai/pipecat-flows/ [3] - https://github.com/pipecat-ai/smart-turn

Disclaimer: I spend a lot of my time working on Pipecat. Also writing about both voice AI in general and Pipecat in particular. For example: https://voiceaiandvoiceagents.com/

I developed a stack on Cloudflare workers where latency is super low and it is cheap to run at scale thanks to Cloudflare pricing.

Runs at around 50 cents per hour using AssemblyAI or Deepgram as the STT, Gemini Flash as LLM and InWorld.ai as the TTS (for me it’s on par with ElevenLabs and super fast)

Technically yes, twilio has sip trunks.

I was more thinking I could add it to my Asterisk server to honey-pot the spam callers into an infinite time waster cycle.

I’m honestly surprised it hasn’t been more prevalent yet. I still get call centre type spam calls where you can hear all the background noise of the rest of the call centre.