The Korean tagline I wrote for the original Play Store listing translates roughly to: “Even a whisper gets translated, reading YouTube by ear.” It was a promise about one specific feature, not a slogan I picked for how it sounded.
What “Listening to Its Own Speaker” Actually Means
Most translation apps expect you to talk into them, or hold them up to someone else talking. TransLee has always done something slightly different: it can also just sit next to whatever sound is already in the room, a phone playing a YouTube video, a TV, a radio, and translate that sound live, without any screen-sharing or app integration on the source side.
There’s no special API doing this. It’s the browser’s microphone, picking up whatever the room’s speaker is putting out, the same way your own ear would. If the source is quiet, the translation gets harder. If it’s clear, it works about as well as someone speaking to the phone directly.
I tested this against a Kenyan news broadcast and an Al Jazeera segment, both playing at a normal, unremarkable speaker volume, not loud, not deliberately quiet. Both got picked up and translated. I haven’t tested an actual human whisper yet, so I won’t claim the tagline is literally true down to the decibel. What I can say is that it handles ordinary broadcast audio at normal volume without needing to be right next to the speaker.
Why This Feature Existed Before Gemini Live Translate
This isn’t new to the Gemini version. The original Flutter app had it too, internally called “Audio Direct mode,” always on, with no toggle exposed in the UI, mostly because it was a feature aimed at a fairly specific use case: understanding a foreign-language video while using Android’s split-screen mode, TransLee running in one half of the screen, YouTube in the other.
Carrying that same behavior into a web app, with no native Android APIs to lean on, meant leaning entirely on the browser’s own microphone permission and playing with how the browser and the operating system handle acoustic echo cancellation, the setting that decides whether a device tries to filter out sound the device itself is producing. Turn that setting on and it can filter out the very audio you’re trying to hear. Turn it off and it works, but it opens a different problem I ran into a day later: if the app is also supposed to speak a translation out loud, that spoken audio can loop right back into the mic as if it were more input.
That collision, one feature quietly breaking another, is its own story. It’s the next post in this series.
How to Actually Use It
Open TransLee, allow microphone access, and pick the default mode, listed as speaker sound translation. Place the phone near whatever is producing the sound you want translated. Text and translated voice both stream back as it plays. Switch to conversation mode only when you’re talking to TransLee directly and want it to speak a translation back to a person in the room.