Chert - Vapi for FaceTime: AI video agents in a few lines
by•
Chert is Vapi for FaceTime. Build and deploy interactive AI video agents that can answer and place FaceTime calls with just a few lines of code.
Deploy agents for remote support, field service, telehealth intake, guided onboarding, or anything that's easier to show than explain.
Try it live: FaceTime an agent right now and show it something.
Replies
Chert
Hey everyone! I'm Gary, co-founder of Chert (YC P26).
Chert is launching Vapi for FaceTime. In a few lines of code, you can deploy an AI agent that answers and places FaceTime calls, sees what the user shows their camera, and responds in real time.
Here's why we're fired up about this: every voice agent shipping today is blind. It can only handle what a customer can describe, so the second a problem is visual (e.g. "which cable goes where?", "what's this error on my screen?"), a human has to jump in. The most natural fix is video.
Now any builder can put an agent on the single most trusted and already-installed video channel: FaceTime. Remote support that can actually see the problem. Telehealth intake that reads the room. Onboarding that walks you through it live. This is the interface AI has been missing.
Try it right now: FaceTime +1 310 279 2297 and check out https://www.trychert.com/facetime
Would love any feedback or thoughts!
Happycapy
@garygao This is a really compelling use of multimodal AI. Voice agents are useful, but letting the agent actually see the problem changes the game, especially for troubleshooting, telehealth, and onboarding. 👍
Chert
@victoria_wu Thank you!
FaceTime is pretty US/Apple-centric, isn't WhatsApp the bigger market globally?
Chert
@arun_pratap_singh7 Yes, we have plans to expand to WhatsApp messaging + video calling in the future!
Can this be used for KYC or identity verification on a video call? Feels like a pretty cool and interesting use case
Chert
@abhi_singh4000 Yes, visual identity verification could actually be more robust than a lot of the current identity verification infrastructures over text, email, or calls today
What are the use cases of this?
Chert
@vimal_singh6 Use cases could be anything from remote support to field service, telehealth intake, or guided onboarding!
How do you handle privacy and consent for the video stream, especially for regulated use cases?
Chert
@rajesh_gupta__ We encrypt user information so that it is never exposed
@garygao Congrats on the launch! The voice agents are blind framing instantly clicked for me.
Chert
@suyash_kr Thanks!!
The latency on the live demo is better than I expected, really impressed!
Chert
@maurya_abhiranjan Thank you!
Humalike
congrats on the launch!
Chert
@mcarmonas Thanks!
Impressive. Do you use TTS or STS model?
Chert
@konstantin_tikhaev Both. We have TTS and Realtime models available
@garygao will test realtime endpoints first to check latency. docs and api keys directly in dashboard?
Refocus
Choosing FaceTime as the surface is the part I keep thinking about. Most voice agents ship on a phone number or an in-app widget, and both carry a tax: the number reads as spam, the widget needs a download.
I build voice AI that calls aging parents at Callie Care, and the constraint that governs everything is that the person on the other end will not install anything or learn a new interface. FaceTime is one of the few places where that is already solved, so the "sees what the user shows their camera" part is more interesting to me than the code-in-a-few-lines part. Half of what an adult child asks over the phone is really "show me the pill bottle."
Gary, two questions. On an agent-placed call, what does it look like on the recipient's lock screen, a normal contact or something visibly automated? And where does your round trip latency land on a live call, since past roughly 800ms people start talking over the agent and the whole illusion drops.
Chert
@igorgurovich I completely agree! A phone number reads as spam, a widget needs a download, and FaceTime skips both. That's why we started there.
Regarding your two questions:
Lock screen: It appears as a normal FaceTime call from the Apple Account/phone identity. If saved, the recipient sees the contact name/photo. Otherwise, just the identifier.
Latency: We offer realtime speech-to-speech right now through OpenAI Realtime and LiveKit. This lets us achieve sub 800ms latency.
Elder care is one of the use cases we're most excited about, and I'd love to have a longer conversation about this. Here's my calendar: https://cal.com/team/chert/chert-call