Reading Google Meet captions without a bot: how it works
There are two ways to hear a call. One is to join it as a participant, a bot that shows up in the list and listens. The other is to read what the page already shows you.
Axelize, the sales assistant I build, takes the second route. It follows a Google Meet call from the captions on screen. No bot joins. No audio or video gets recorded. That one choice decides most of what the interface looks like, so here’s the chain from caption to reply, step by step.
How a call becomes a reply
Step 1
You turn captions on
Google’s own steps are short. Join a meeting, hover over Captions and translation at the bottom, and choose Turn on captions. Pressing c does the same. Google also notes that “The captions are turned on only for you”, so nobody else on the call sees anything change.
Step 2
The extension reads the lines
A script that runs inside the Meet tab reads the caption text as Meet draws it. That’s the whole listening layer. There’s no audio stream to capture and nothing to join.
Step 3
The lines go to Axelize
The extension passes them on in small batches. Axelize reads the conversation so far, along with what it knows about the person you’re talking to, and writes your next sentence.
Step 4
You read it from the side panel
The reply appears in Chrome’s side panel, next to the call, written in your voice and ready to say. You never leave the meeting to find it.
Why captions and not audio
A bot changes the call. The other person sees a new name, and some of them will ask what it is. Recording audio raises a second set of questions about consent and storage.
Reading captions avoids both. The call looks the same to the buyer. Same names in the list, same window, nothing new to explain. The only thing the extension handles is text you already chose to turn on. That’s a product choice, but it’s also a trust choice, and trust is a design problem like any other.
“No bot” isn’t the same promise as “no audio”, and neither is the same as “no transcript”. We read seven live assistants’ own privacy pages and wrote up what each one stores. It’s worth reading before you pick a tool for your team.
What it means for the interface
Working from captions shapes the panel in two ways.
Text arrives line by line, so there’s no recording to wait for. The reply can be written while the buyer is still talking.
And because the input is plain text, the output can be plain too. One short sentence you can read without scanning for it.
Axelize works with Google Meet. Zoom and Microsoft Teams are on the roadmap.