No, not the band.

I would enjoy the following UX:

Each smart phone has an app. The app listens to your speech..., it uses the really-good STT systems that make auto-subtitles e.g. on google chat. Don't be fooled by youtube. Stuff already exists.

Especially if it could be "on the edge", which I think we want anyhow since it's closest to the source. Maybe, should it come up, "the other end" could finish a partially-completed transcription, depending on latency or something.

In any case, you talk, a "text chat UI" updates without asking you and sends your text message transcribed from the speech. On the other end, exactly the same thing happens.

So you read a chat but respond causally with your voice, being unafraid of mistranscriptions, which do not happen, and so does everyone else, as a gimmick, like "only X characters" and other such.