What Is Veo 4? What Google Has Confirmed So Far

I was watching Google I/O 2026 the way a lot of people in video-gen work were watching: waiting to see if what is veo 4 would finally get an answer on stage. It didn’t. Two days of keynotes, dozens of announcements, and the name “Veo 4” was not among them. That absence is the most useful piece of information anyone planning around Google’s video stack got this month.

Dora here. This piece is a status check, not a roadmap. I’m going to separate what Google has actually confirmed from what the internet is still calling “Veo 4,” and then talk about what teams shipping right now should actually do with that information.

Current Status: Veo 4 Is Not a Product Google Has Announced

As of late May 2026, there is no Google announcement, product page, model card, API spec, or pricing sheet for a model called Veo 4. The official Veo model line on Google DeepMind lists Veo 3.1 as the current model. The Google I/O 2026 announcements collection does not mention Veo 4 once.

What Google did announce on May 19 is Gemini Omni — and that detail reshapes the question entirely. More on that in a minute.

So the short answer to “what is Veo 4”: it is, today, a rumor and a search query. Not a product.

What We Know for Certain

A few things are solid.

Veo 3.1 is the current shipping model. It was released on October 15, 2025, and it’s documented, accessible through Vertex AI and the Gemini app, and priced. The capabilities are concrete: 1080p output, native audio generation, 8-second clips that can be chained, and ingredients-based scene building. This is the model in production today.

Google’s historical release pattern is real but not a guarantee. Veo 1 launched at I/O in May 2024. Veo 3 launched at I/O in May 2025. That pattern is why so many people expected a Veo 4 announcement in May 2026. According to Wikipedia’s Veo entry, Veo 3.1 was released off-cycle in October 2025, which already broke the I/O-only cadence. Two data points are a coincidence. Three is a pattern. We had two.

Google chose a different headline. Instead of Veo 4, Sundar Pichai’s I/O 2026 keynote introduced Gemini Omni — described as a model that can “create anything from any input, starting with video.” Omni is not Veo 4 renamed. It’s a different architecture: Gemini’s reasoning layer fused with the generative media stack that includes Veo, Nano Banana, and Genie.

That last point is the one I keep coming back to.

Rumored Veo 4 Features — and Why They Stayed Rumors

For the months leading up to I/O 2026, the veo 4 features rumor list was consistent across leak trackers and analyst threads:

4K native resolution
30-second clips
Improved character consistency across cuts
Tighter lip-sync and multi-language audio
Lower per-second cost

None of these were confirmed. None of them were denied either. What happened instead is that Google announced a product that absorbs some of these capabilities into a different surface. Gemini Omni’s pitch is conversational video editing, world-physics understanding, and multi-input fusion — not “Veo 3.1 plus more pixels.”

So the veo 4 announcement people were waiting for didn’t happen, and the gap got filled by something that wasn’t on most rumor sheets.

This is where I’d pause. The question stops being “when does Veo 4 ship” and becomes “is Veo 4 still the right thing to wait for, or did Google quietly route around it?”

I don’t know the answer. Neither does anyone outside of Google DeepMind. But the veo 4 release date question is no longer the operative one for anyone making a build-versus-wait decision.

What’s Not Confirmed — And Why That Matters

Here’s the unconfirmed list in one place, so nobody has to guess:

Question	Status
Is “Veo 4” the next model name?	Unconfirmed. Could be renamed, merged into Omni, or skipped.
Release date	None announced
Resolution / clip length	No spec published
API surface	Unknown (Vertex AI? Gemini API? Different?)
Pricing	Not disclosed
Whether a standalone Veo 4 exists in development	No public statement

I want to be specific about that last row. google deepmind veo is an active product line — that’s documented. Whether the next iteration ships as “Veo 4” or gets absorbed into the Omni family is the part nobody outside Google can answer.

The Omni Pivot Changes the Question

This is the part of the article most “Veo 4 explainers” online are skipping, and I think it’s the most important thing to understand.

For the last three years, Google’s generative media strategy was split: Veo for video, Imagen for images, Nano Banana for image editing, Genie for simulations. Each was its own model line. The expected progression was Veo 3 → Veo 3.1 → Veo 4 → Veo 5.

Omni breaks that progression. According to the official Gemini Omni announcement, Omni is Gemini’s reasoning layer applied to generative media — meaning the video output is generated by a system that also understands physics, history, and intent across multi-turn edits. It’s not framed as a Veo replacement. But it’s also not framed as a Veo successor.

What this might mean — and I’m being careful with “might”:

Veo 4 could still ship, as a dedicated video model alongside Omni, the way Veo 3.1 sits alongside Imagen 4
Veo 4 could be merged into Omni — meaning the capabilities people expected from Veo 4 (longer clips, 4K, better consistency) show up under the Omni name instead
Veo could continue on a slower track, with Omni as the consumer-facing front door

Google hasn’t said which of these is happening. I checked the I/O materials, the DeepMind blog, and the Vertex AI changelog. No statement either way.

That ambiguity is the actual answer to google veo 4 as a question today.

What to Do If You’re Building Right Now

This is where I get practical, because I think the rumor-watching has gone on long enough.

If you need video generation in production in Q2 or Q3 2026, Veo 3.1 is documented, stable, and on Vertex AI. Kling 3.0 and Runway Gen-4.5 ship the longer-clip and higher-resolution work that Veo 4 was rumored to deliver. Ship with what exists. Switch later if something better lands.

If you’re in R&D and the decision is “do we wait six more months for Veo 4,” I’d flip it: assume Veo 4 doesn’t ship under that name, or ships in a form different from the rumor list. Build a stack that can swap model providers without a rewrite. If Veo 4 does land and is better, switching is cheap. If it doesn’t, you’ve already shipped.

If you’re competitive — meaning your competitors are publishing AI video weekly — they’re not waiting. They’re on Veo 3.1, Kling, Runway, or some combination. The cost of waiting on a model that may not exist under the name you expect is six months of compounding output loss.

That’s the framework. The rest is your call.

FAQ

Is Veo 4 officially announced by Google? No. As of May 26, 2026, there has been no Google statement, product page, model card, or pricing announcement for Veo 4. Google I/O 2026 came and went without it.

When will Veo 4 be available? Unknown. Google has not published a timeline. The historical I/O pattern broke when Veo 3.1 shipped in October 2025, so the May-launch precedent is no longer reliable.

What will Veo 4 cost? Unannounced. There’s no pricing information because there’s no announced product.

Can I get early access to Veo 4? Not publicly. There is no early access program, waitlist, or research preview for a model called Veo 4.

Should I wait for Veo 4 or use Veo 3.1 now? If you have a production timeline in the next two quarters, use what’s available — Veo 3.1, Kling 3.0, or Runway Gen-4.5. Waiting for an unannounced model with no confirmed feature set is a higher-risk bet than shipping with documented tools and swapping later.

What’s the difference between Veo 3.1 and the rumored Veo 4? Veo 3.1 is real: 1080p, native audio, 8-second chainable clips, documented on Vertex AI. The rumored Veo 4 list — 4K, 30-second clips, better consistency — is internet speculation. None of it is on a Google product page.

Will Veo 4 be on Vertex AI like Veo 3.1? If it ships under that name, the most likely surface is Vertex AI plus the Gemini API, since that’s where Veo 3.1 lives. But “likely” is not “confirmed.” The Omni rollout uses the Gemini app and Google Flow as primary surfaces, which is a different distribution pattern.

What I’m Watching Next

I’m not going to predict a date. What I will do is watch three signals: a Vertex AI changelog entry mentioning Veo 4, a DeepMind research paper with that name, or a Google Cloud product page. Until one of those exists, what is veo 4 has the same answer it has today.

Previous posts：