Android 17 brings Gemini built into the operating system

At this year's Google I/O, Google introduced Android 17 with a change that alters less how the system looks and more how it's built: Gemini stops being an app you open and becomes a layer of the operating system itself. According to Google, the model — called Gemini Omni in this integration — is available at the system level, reachable by any app, without you having to leave the context you're in.

Alongside it, Google announced two more pieces built into the system. Lyria 3 generates music from text inside Android itself, with no separate app. And AudioLM does real-time translation running on the device — what Google describes as on-device processing, without sending the audio to the cloud.

What changes when AI becomes a system layer

Until now, using AI on your phone meant opening an app, pasting in some text, copying the answer back. Every task carried that friction of stepping in and out. By putting Gemini at the system level, Google removes the middleman: the AI is now where you already are — in the keyboard, the camera, the text selection, in any app.

The difference is one of kind, not degree. An app is something you choose to open. A system layer is something that's always present, available for any other piece of software to call on. For anyone building product, that reframes the question: it's no longer "should I put AI in my app?" but "what do I do with an AI that's already on everyone's device?".

On-device: lower latency and more privacy

The most relevant technical point in the announcement is AudioLM's on-device processing. Running translation on the device itself, with no round trip to a server, has two direct consequences.

The first is latency. Real-time translation only works if it is, in fact, real time — and every trip to the cloud adds delay. Processing locally cuts that gap.

The second is privacy. If the audio never leaves the device, there's no transmission of a conversation to a remote data center. For translating personal conversations, meetings, or customer service, that reduces the data's exposure surface. It's an architecture choice with a practical effect for anyone who cares about where their data travels.

What changes for those who create product and content

With Lyria 3 generating music from text inside the system, and Gemini available in any app, the cost of experimenting with generated media drops. A scratch track for a video, a live-translated caption, writing assistance inside the app you already use — none of it requires an extra tool anymore.

For content creators, the consequence isn't that the machine starts creating in your place. It's that the barrier to entry for drafting, testing, and iterating goes down. The first version gets cheaper to produce. The work of deciding what holds up, what has a purpose, and what actually ships stays human — and becomes more visible, because the mechanical part drops out of the way.

Worth stating plainly: Google announced these features at I/O. Actual availability, the languages AudioLM supports, and Lyria 3's quality outside a demo are things only production use will confirm. A stage announcement and day-to-day operation don't always line up.

Ubiquity isn't an advantage — knowing where to apply it is

Here's the point that matters to us at 10Dobro. When an AI tool is on every Android device, it stops being a competitive edge. If it's available to everyone, no one gains an advantage just by having it. Gemini in the operating system is the floor, not the ceiling.

The advantage moves elsewhere: knowing where to apply the tool, at which step of the process, by what criteria. A ubiquitous AI layer solves the "how" — generate, translate, draft. It doesn't solve the "where" or the "why." Those remain the work of whoever understands the client's problem, the context of the operation, and the result that needs to come out.

It's the thesis we work from: AI doesn't replace teams, it multiplies what a good team already delivers. That's how we reached 26 systems in operation alongside 150+ audiovisual productions — combining automation with people who know where it makes a difference. Android 17 puts a good tool in everyone's hands. What you do with it, and at what point in your operation it pays off, is still the part that decides.