Pick up every call, in every language.
In-browser voice calls between you and your visitors, with full transcription. AI voice answers when nobody's around, books the meeting, and translates live for the rest of the conversation.
Maria, live call
02:47, recording
Live transcript, auto translated
Maria at 00:14
“Could you walk me through the deployment options for a regulated stack?”
Hamza, agent at 00:22
“Yes. We can deploy on a private VM, your VPC, or shared infra. Which fits your compliance bar?”
Maria at 00:31
When chat is not enough, go voice. With a human or with your AI.
Either side taps to call. The browser handles the audio with echo cancellation, noise suppression, and automatic gain. The visitor can also call your AI sales agent directly. Powered by Gemini Live, it talks back in the same voice you picked, fires tools mid call (lead form, payment links), and barge in cuts its reply the moment the visitor speaks. Every word lands in the chat thread. Recordings are stored as a stereo waveform you can scrub.
AI voice agent, picked per site
Choose Puck, Charon, Fenrir, Kore, or Aoede. The bot greets in the visitor language and tools fire mid call exactly like text chat.
Live transcript in chat
Server side speech to text through Gemini, written into the same conversation thread as the text bubbles.
Recorded as a stereo waveform
Visitor on the left channel, agent or AI on the right. Stored in your Google Cloud bucket, played back through a 40 bar scrubber bubble.
Calls hand off cleanly
Status bubble flips visibly when a call ends so chat continues without a context break. AI calls auto answer after the response delay you set.
Then email lands in the same inbox.
Customer wrote in instead of called? It threads next to the chat. See email helpdesk.