Google quietly launched an offline-first dictation app on iOS called Google AI Edge Eloquent.
The interesting part isn’t “dictation.” It’s the architecture. Once you download the models, the speech-to-text runs on-device using Gemma-based ASR, so it can work without a connection.
After you pause, it cleans up what you said by removing filler words like “um” and “ah,” then gives you one-tap transforms such as Key points, Formal, Short, and Long.
And Google left a clear escape hatch: you can keep it local-only, or flip on cloud mode where Gemini helps with text cleanup.
It even supports importing names/jargon (and adding custom words), which is exactly what makes dictation usable in real work.
Takeaway: offline AI is becoming the default because it’s faster, more private by design, and doesn’t fall apart when connectivity does.
If you’re thinking about this internally, happy to compare notes.
Google quietly launched an offline-first dictation app on iOS called Google AI Edge Eloquent.
