That's cool. I still think the primary issue is the marketing makes it sound like the ear buds are bringing new tech to make this experience possible, when they're really bringing:

a) UX for "hold while talking"

b) Known quality of the headset

From the outside, it looks a bit shady, and a bit un-Googlelike just to compete with Apple.

I imagine a more Google-like approach would've been to release Translate as a local model (eg like https://developers.google.com/vision/). It's easy to imagine the community would quickly figure out a way to solve a) and make the experience work for most headsets.