As sister comments have pointed out, I find that it's very much dependent on the phone you're using. If it can't even show me a waveform then I don't think it's going to be able to tease out any sort of speech

I was using Google's Translate app.

Sorry I wasn't clear. It's about the hardware microphone. They're usually optimized to filter out far field sounds and only focus on near field sounds. So if you're trying to understand a conversation from even 10 ft away, it can be a problem. Not saying that's the situation that you encountered, but it does make sense that a $50 phone versus a $500 phone might have different capabilities when it comes to the microphone

I'm talking about Gemini Live in the Gemini App, not Google Translate.

Tell Gemini Live which speaker you want to engage with, and it is relatively intelligent about it.

"Only worked in a 1:1 in a quiet place"

well then its not model problem

It's not a model problem, but it is a SW problem. It should be able to distinguish nearby people from people farther away and give me options to set a threshold on who to include.