Google I/O 2026 preview — Gemini Omni video model surfaces in a UI leak; what to actually watch tomorrow

The Omni leak — and why it matters more than the headline suggests

On May 11, 2026 — one week before I/O — a model-card UI string for Gemini Omni surfaced on Reddit. The visible text, confirmed by Testing Catalog and corroborated across multiple secondary outlets:

“Create with Gemini Omni: meet our new video model, remix your videos, edit directly in chat, try templates, and more.”

The leak appears to have been either accidental or a limited A/B test of pre-release UI. A “usage limits” tab appeared in the same UI window, with early testers reporting credit consumption — strongly suggesting the model is in late-stage staging rather than concept work. Early outputs from the limited surface showed Omni:

• Excellent at video editing operations — watermark removal, object replacement, scene rewriting via text instructions all worked at unusually high quality for a first public glimpse
• Behind ByteDance’s Seedance 2 on raw generation fidelity — the gap is real but not catastrophic
• Likely shipping in tiered variants (Flash + Pro), consistent with Google’s standard product packaging

The strategic framing — and the part most coverage will get wrong — is that Omni isn’t about beating Seedance 2 or Sora. It’s the Nano Banana Pro playbook applied to video: not the highest-quality model in raw benchmark testing, but the model most casual users will actually have access to because it’s free or near-free inside Gemini.

If Omni ships at the Gemini free tier on May 19, every Gemini user gets a “good-enough” AI video generator in their existing AI subscription. That distribution moat matters more than the 5-10% fidelity gap with the cutting edge. The same dynamic played out in image generation — Nano Banana Pro shipped at free / low cost inside Gemini, and the Midjourney $30/month lock-in eroded fast for casual users. The video category may follow the same pattern.

What to watch at I/O 2026 (beyond Omni)

Several other threads will likely surface during the keynote tomorrow:

Gemini 3.5 (rumored)

Industry expectation is that Google will announce a meaningful upgrade to the flagship Gemini line — Gemini 3.5 — at I/O. The current Gemini 3 Pro / Flash has been the workhorse for Gemini consumer surfaces; a 3.5-class upgrade would tighten the Claude Opus 4.7 and GPT-5.5 competitive gap. What to watch: benchmark deltas on AIME, GPQA, SWE-Bench, and MMMU — and pricing if API tiers change.

Android 17 + Gemini Intelligence integration depth

Google has been publicly framing Android 17 as a “Gemini-first” operating system. Per CNBC coverage on May 12, Gemini Intelligence will be able to move across apps, understand what’s on the screen, and complete tasks that previously required jumping between services — for example asking Gemini to look at a guest list, build a menu, and add ingredients to your shopping cart, all without leaving the conversation.

What to watch: which apps actually have deep agent-layer hooks at launch, whether Apple’s WWDC counter-move (typically early June) is already factored in, and how the Pixel feature drop differs from the broader Android rollout.

Aluminium OS

A persistent rumor — covered across Android Authority and other pre-I/O analyses — that Google will unveil a new OS called Aluminium, possibly an evolution or rebrand of ChromeOS aimed at unifying ChromeOS, Android tablet, and emerging form factors. Treat this as speculative until announced; if it ships, it materially changes Google’s hardware roadmap.

Agent-layer cross-app actions

Both Anthropic (May 5 finance-agent push) and OpenAI (May 11 Deployment Company $4B services arm) have been moving aggressively on agent-layer enterprise deployment. Google’s response will likely be agents that run inside the Google ecosystem — across Gmail, Calendar, Drive, Workspace, Search, Maps — leveraging Google’s unique advantage of owning the underlying data graph. What to watch: agent capabilities specifically tied to Workspace + Gmail context that competitors can’t replicate.

Workspace integration deepening

Anthropic’s Microsoft 365 GA in early May was a direct attack on the Workspace flank where Google has had a structural advantage. Expect Google to push back hard on the I/O stage with Workspace-native Gemini features that match or exceed what Claude offers inside M365.

Android XR + Gemini

Google’s XR (extended reality) push continues. The relevant question: are AI features (Gemini-on-glasses) credible at I/O, or still demo-stage? Apple’s Vision Pro hasn’t broken into mainstream; the AR/VR + AI hardware story is still unsettled.

What it means for Pick Right readers tracking video AI

The AI video category currently lines up as:

• Veo 3 — Google’s existing flagship; strong synced audio capability; integrated into Flow editing tool
• Runway — Western quality leader; just-disclosed $5.3B valuation and +$40M ARR in Q2; strong creative-workflow tooling
• Kling — Chinese price-performance leader; long-duration clips; aggressive pricing
• Sora — shut down April 26, 2026; OpenAI exited the consumer video product
• ByteDance Seedance 2 — current quality leader on raw fidelity per pre-I/O coverage
• Gemini Omni (pending) — Google’s incoming play

If Omni lands at the Gemini free tier tomorrow, the casual-creator video tool conversation shifts. Pick Right’s standing recommendation that Veo 3 is the better casual-user pick than Runway for users already in Google’s ecosystem stands; Omni would extend that pattern further.

If Omni ships as a paid Gemini AI Pro feature only, the dynamic is closer to a Runway competitor with Google’s distribution muscle behind it — meaningful but not category-redefining.

The pragmatic read on which scenario is more likely: Google’s recent product packaging — Nano Banana Pro at the free tier, Veo with limited free access — suggests Omni will have a meaningfully accessible free or low-tier surface. Google’s strategic priority is distribution over per-unit monetization on the consumer-creative-AI axis; that pattern argues for broad Omni availability.

What this means for Pick Right reviews

Tomorrow’s loop run will publish the I/O recap article with whatever actually ships — Omni pricing and capabilities, Gemini 3.5 if announced, the agent-layer features, and any pricing/tier changes to existing Gemini subscriptions. The relevant Pick Right reviews to watch for refresh signals:

• Gemini — model lineup likely changes; pricing tiers may shift
• Veo 3 — if Omni replaces or supplements Veo, this review needs material updates
• Nano Banana Pro — may get repositioned alongside Omni as the “Gemini-native creative” pair
• Kling and Runway — competitive positioning shifts based on what Omni’s pricing looks like

For broader context on the video AI category, see best AI video tools and the head-to-head Runway vs Veo 3 comparison.

The pragmatic read going into tomorrow

I/O keynotes have a long history of announcing more than ships. The pre-event leak signals for Omni are strong (model card visible in production UI, usage-limits tab live, early testers reporting actual generations). The other expectations — Gemini 3.5, Aluminium OS, deep agent-layer features — are softer; some will ship, some will be teased for later in 2026.

For Pick Right readers, the pragmatic stance is: don’t change tool stacks today based on tomorrow’s announcements. If Omni ships and meaningfully changes the casual-creator video math, Pick Right will reflect that in the Veo 3, Runway, and Kling reviews within 24-48 hours of the announcement. The right time to act on I/O is the week after, not the day of.

Watch the keynote tomorrow at 10am PT / 1pm ET. Pick Right’s recap publishes within the day with the structural read — what actually changes, what’s hype, and what (if anything) you should do about it.

Questions or corrections? Email Pick Right. Want the full list? See all news.