EDITORIAL NOTE: This is a live-coverage placeholder. Sections will be filled with verified facts during and after the Google I/O 2026 keynote on Monday, May 19, at 09:00 ET. Every announcement listed below is based on prior leaks and previews. The final published version will carry only confirmed launches.
Google I/O 2026 runs May 19 and May 20. The keynote opens at 09:00 Pacific on Monday. io.google/2026 hosts the official agenda and livestream. Creative AI is the throughline this year: a unified Gemini video model, a Veo successor with native audio, an Imagen rebrand, a Lyria upgrade, an upgraded Project Astra agent, and new Android XR creator tools are all expected on stage.
Here is what creators should brace for, in the order Google is most likely to reveal them. Gemini Omni is the rumored unified video and image model, surfaced as a string inside Gemini's video tab in early May; if real, it consolidates Veo, Imagen, and possibly Lyria behind a single Gemini-native model. Veo 4 is the production video generator, expected to extend shots past eight seconds and add native dialogue plus ambient audio. Imagen 4 (DeepMind codename Spark Robin) is the image upgrade with rumored long-text rendering and multi-reference subject control. Lyria 3 Pro targets longer, structured music generation that competes head-to-head with Suno v5 and Udio v1.5. Project Astra is the always-on agent layer that ties Gemini into Workspace, Android, and the on-device camera. Android XR ships its first creator-grade toolkit on the Samsung-Google reference headset, including an immersive video editor and a 3D background generator. Pricing, tier ceilings, and Workspace integration are the three levers to watch: those decide whether the new models reach creators on the free or Pro tier, whether they land in Docs, Slides, and Meet, and whether Vertex AI undercuts Suno, Runway, ElevenLabs, and Midjourney on per-call cost.
This article will be updated section by section as each reveal lands. Bookmark this page for the running summary, the official URLs, and the workflow integrations creators can start using today.
Gemini Omni: Google's Unified Video Model
The biggest leak ahead of the keynote is Gemini Omni, a string surfaced inside Gemini's video tab in early May. The leak hints at a single Gemini-native model that handles image, video, and possibly audio in one system, replacing or fronting Veo. Our prior coverage walks through the leaked UI strings and the three possible interpretations in Gemini Omni: What the leaked UI string tells us. Expected creator angle: text-to-video, image-to-video, and conversational edit in the same surface, with Workspace and Android distribution alongside.
What 2025 to early 2026 showed: creators want one consolidated surface, not four. Veo 3.1, Imagen 3, Lyria 2, and the Gemini chat tab all live in separate UIs today. A unified model with one prompt box, one history, and one billing meter is the single biggest workflow shift this keynote could deliver.
Veo 4: Native Audio and Longer Shots
Veo is Google's flagship video model. The current production version (Veo 3.1) generates eight-second clips with synchronized audio. Veo 4 is expected to extend shot length, add native dialogue, and ship a Vertex AI tier alongside the consumer Gemini app. Track the product page at deepmind.google/technologies/veo for the official capability sheet once the keynote lands.
What creators want: shot length over thirty seconds, lip-synced dialogue without a separate audio pass, and a price floor that matches OpenAI Sora 2 and Runway Gen-5. The Veo 3.1 ceiling of eight seconds forces multi-shot stitching in every workflow longer than a teaser.
Imagen 4 and the Spark Robin Codename
Imagen 4 (codenamed Spark Robin in DeepMind preview channels) is expected to land alongside Veo 4. The product page at deepmind.google/technologies/imagen still references Imagen 3 as of writing. Watch for native long-text rendering, multi-reference subject control, and a new pricing tier that lines up with Vertex AI image generation. Our preview of the expected releases is at Google I/O 2026 preview: Veo 4 and Imagen 4.
What creators want: reliable text rendering inside the image (the headline weakness of Imagen 3 versus GPT Image 2 and Flux Klein 9B), subject consistency across a batch, and a per-image price under one cent that survives the Qwen-Image-2.0-Pro and Flux pricing floor.
Lyria 3 Pro: Frontier Audio Generation
Lyria is Google's audio model. Lyria 2 shipped in early 2026 with instrument-level prompt control. Lyria 3 Pro is the expected I/O 2026 upgrade, targeting longer compositions, structured song generation, and a Vertex AI music tier that competes with Suno v5 and Udio v1.5. Reference: deepmind.google/technologies/lyria.
What creators want: full-length tracks with verse, chorus, and bridge structure rather than two-minute loops, stem separation on export, and a rights model that clears commercial use on YouTube and TikTok. Suno v5 set the bar at four-minute structured songs in April; Lyria has to match or exceed.
Project Astra: The Always-On Agent
Project Astra is Google's universal agent. The upgrade expected at I/O ties Astra into Workspace, Android, and the Gemini app as a context-aware assistant that can see the screen, hear the room, and act across apps. The product page at deepmind.google/technologies/project-astra will host the live spec. Our coverage of the broader Gemini agent stack is in Google Gemini Spark: Always-On AI Agent Coming to I/O 2026.
What creators want: a stable agent that can act inside Photoshop, Premiere, Figma, and Blender, not just inside Workspace. Anthropic's Claude for Creative Work suite set the benchmark for named-app integrations in April; Astra has to match that surface area or it stays a demo.
Android XR: Creator Tooling on the Headset
Android XR ships its first major creator-facing toolkit at I/O. Expected: a Veo-powered immersive video editor, an Imagen-driven 3D background generator, and an Astra developer SDK that runs on the Samsung-Google reference headset. The platform page at blog.google/products/android/android-xr will carry the launch post. Companion context for the Android creator stack lives in Android Show 2026: Googlebook, Vibe Widgets, Premiere App.
What creators want: a workflow that does not require buying the headset on day one. A web-based preview surface, an export path to standard 2D files, and an SDK that runs on existing Android dev kits would let the toolkit reach the wider creator base instead of the few thousand headset owners.
What to Watch During the Keynote
Three reads on the keynote will tell you which announcements stick. First: which Gemini tier gets which model. A Pro-only Veo 4 ceiling kneecaps creators on the free tier; a Plus-tier ceiling brings parity with ChatGPT Plus. Second: which Workspace surface gets the new image and video models. Native Docs, Slides, and Meet integration is the largest single distribution lever Google can pull. Third: which Vertex AI prices land. Suno, Runway, ElevenLabs, and Midjourney all compete on per-call cost; Google's pricing sheet at deepmind.google/models/gemini sets the floor for the rest of the market.
Frequently asked questions
When does Google I/O 2026 start?
The opening keynote runs Monday, May 19, at 09:00 Pacific. Day one focuses on product reveals. Day two is the developer keynote with Android, Vertex AI, and Workspace deep dives.
How can I watch live?
The livestream is on the official I/O site at io.google/2026 and on the Google YouTube channel. The on-demand recording lands within the hour.
Will Veo 4 be free?
The free tier is unconfirmed. Veo 3.1 lives on the Pro and Workspace tiers today. Expect Veo 4 to slot in at the same ceiling, with a watermarked free preview as a possibility.
What is Gemini Omni in plain terms?
The leaked product name for a Gemini-native model that handles image, video, and possibly audio in one system. The naming may be product-only, with Veo and Imagen still doing the generation work behind the scenes.
Where will the official model cards land?
Watch deepmind.google/models/gemini and the individual Veo, Imagen, Lyria, and Astra product pages on deepmind.google/technologies. Each lands a refreshed spec sheet within minutes of the keynote.
