Android 17 Arrives on Pixel with Gemini Omni, Lyria 3, and a New Multitasking Interface

Google released Android 17 to Pixel devices on Tuesday, June 16, alongside its monthly Pixel Drop feature batch. The update is the most substantive AI integration in Android’s history — not in terms of a single landmark feature, but in the sheer density of AI-powered capabilities landing across the platform simultaneously.

The timing is deliberate. Apple’s WWDC in June 2026 showcased iOS 27 with deeper Gemini integration (following Apple’s partnership with Google announced at WWDC 2025). Android 17 is Google’s answer: a demonstration that the original Android platform can ship the actual AI features, not just promise them.

Gemini Omni: AI in the Editing Process

The centerpiece AI addition in Android 17 is Gemini Omni, a multimodal model that integrates video editing directly into conversational interactions within the Gemini app.

The premise is straightforward but the execution is genuinely new. Rather than opening a separate video editing application, users can share a video with Gemini Omni and describe the edits they want in natural language — trim this section, change the color temperature, add a text overlay, remove the background audio in the first fifteen seconds. The model processes both the video content and the instruction, makes the edit, and returns the modified clip within the conversation.

This is qualitatively different from the “AI-assisted editing” features that most video apps have been shipping since 2024. Those features present AI as a button within an editing interface. Gemini Omni makes the conversation the interface. The editing happens inside the chat window, which means the interaction model is the same whether you’re asking Gemini to schedule a meeting, write an email, or trim a video clip.

The feature ships first on Pixel 8 through 10a and requires the Gemini app. It is a Gemini Pro-tier exclusive at launch, meaning free Gemini users will see a prompt to upgrade before they can access Omni editing.

Lyria 3: Music Generation Arrives on Android

Google’s AI music generation model Lyria 3 also arrives in Android 17, integrated directly into the Gemini app on Pixel devices. The feature allows users to describe a piece of music — including style, mood, tempo, instrumentation, and whether they want vocals — and generate a custom track from the prompt.

Lyria 3 also accepts image inputs. Users can photograph a scene — a sunset, a crowded street, an empty room — and the model will generate a music track it interprets as fitting the visual. This dual-input modality (text plus image) is designed to make music generation useful for content creators building short videos, where the goal is not to compose music from musical knowledge but to find audio that fits a visual mood.

The generation quality on Lyria 3 is a significant step up from Lyria 2, which launched in mid-2025. The new model handles transitions between sections more coherently, produces more structurally varied compositions, and is substantially better at following tempo and instrumentation instructions. It remains a tool for generating background music and ambient audio rather than a replacement for professional composition, but the gap to “good enough for consumer content creation” has closed considerably.

AudioLM on Pixel 10a: Speech Translation That Actually Works

The Pixel 10a, released earlier this year as Google’s mid-range flagship, receives a differentiated capability in Android 17: AudioLM-powered speech-to-speech translation.

AudioLM is Google DeepMind’s audio language model, which processes and generates audio directly without routing through text as an intermediate representation. Applied to translation, this means the system can capture the prosodic qualities of speech — rhythm, emphasis, pace, and some emotional tone — rather than just the words, producing translated output that sounds more like a natural speaker than a text-to-speech system reading a transcript.

In practical terms, the Pixel 10a can now serve as a near-real-time translation device for conversations. One person speaks in their native language; the phone translates and speaks the output in the target language with natural-sounding delivery. The system works offline for the 25 most common language pairs, with cloud-assisted translation available for an additional 75+ pairs when connected.

This is not the first phone to offer live translation — Google Translate has had Live Transcribe for years, and the Pixel 6 introduced Interpreter Mode. But AudioLM represents a generational improvement in the naturalness of translated speech output, addressing the core UX problem that made previous translation tools feel awkward in real conversations.

The Bubble Bar: Multitasking Gets a Redesign

Beyond the AI-specific features, Android 17 introduces the Bubble Bar — a fundamental change to the multitasking interface.

The Bubble Bar replaces Android’s traditional recent-apps view for certain interaction patterns. Active apps and conversations can be pinned as floating bubbles along the bottom edge of the screen, allowing users to switch between them without leaving the current context. A long-press on any bubble expands a persistent widget-style view of that app’s current state, while a tap brings it full-screen.

The design is similar in concept to the chat head feature Android introduced years ago for messaging apps, but extended across all application categories. A navigation app can stay as a bubble providing turn-by-turn audio while a user browses content in the foreground. A music player bubble shows playback controls without the user leaving their current app. Multiple conversation threads — across different messaging apps — can coexist as bubbles.

Android 17 also adds screen reaction recording (capturing the selfie camera and device screen simultaneously for reaction content), a foldable gaming mode with a 50/50 split layout and dynamically resizing game pad, and live app mirroring from Pixel phones to Pixel Watch.

The Gemini Intelligence Gap

The one conspicuous absence in Android 17 is the feature Google most prominently teased: Gemini Intelligence, the agentic assistant layer that would allow Gemini to take multi-step actions across apps, make purchases, book appointments, and execute complex workflows autonomously.

Gemini Intelligence was previewed at Google I/O 2026 as the next evolution of the Google Assistant model — a system that could understand long-horizon tasks like “plan a trip to Tokyo in October, book a hotel near Shibuya, add it to my calendar, and send the itinerary to my travel companion” as single instructions. The capability represents the genuinely transformative version of on-device AI: an assistant that does things, not just answers questions.

That feature is not shipping in Android 17. Existing Pixel owners who watched Google I/O expecting to receive this capability will find it absent from the update notes. A Google spokesperson confirmed that Gemini Intelligence “will roll out to supported Pixel devices in a future update,” without specifying a timeline.

The gap between what Google previews at I/O and what ships in Pixel updates has been a recurring critique of Google’s AI strategy — the company has consistently demonstrated features months before they are available to users. Android 17’s AI features are real and meaningful, but the feature that would make Gemini a genuine competitor to what Apple described for iOS 27 remains in preview.

For now, what Android users have is a platform that generates music, edits videos in conversation, translates speech more naturally than before, and reorganizes multitasking around persistent app bubbles. That’s a significant leap from Android 16. The next leap — autonomous action — is still coming.