On-device by design

The AI runs on your phone. Not on our server.

Three reasons we made this choice — privacy, latency, reliability. One reason we stand by it — Gemma 4 E2B finally makes it possible without compromise.

  • 100% 100% on-device inference
  • 0 0 user content uploaded
  • 1× pay-once purchase

Your travel notes never leave your phone

Every word you translate, every photo you analyze, every voice memo you record — Cove processes all of it locally. There is no cloud round-trip. There is no log on a server you cannot audit. The privacy boundary is the device boundary, which is the only boundary that actually holds.

→ Read the full network-call audit

Real measured latency, not best-case marketing

Gemma 4 E2B running with NPU acceleration on a Pixel 9 returns a translation in under 500 ms — every time, regardless of network. A typical cloud round-trip on a flaky travel connection is 2–8 seconds. The difference is what makes a back-and-forth conversation possible at the train counter.

Airplane mode, captive Wi-Fi, train tunnels

Travel is network-hostile. Subway tunnels, basement restaurants, captive Wi-Fi portals you cannot get past, eSIMs that promise 4G and deliver 2 kbps. Cove keeps working in all of them — because nothing it needs lives on a server.

What we trade away to get this

On-device is not free. The first download is 2.5 GB. Long professional documents still benefit from cloud-scale models like DeepL. Very rare languages are a real gap. We list these openly because the case for on-device is strong without exaggerating it.

Same architecture across the whole family

Travel is the first to ship, but the on-device promise is the same for Voice, Photo, and Health when they go live: every byte of your content stays on your phone.

Ready to try it?

Cove Travel is the first product. One purchase. No subscription. Buy once, keep forever.