Icon

Updated weekly

OW

OW

Icon

Updated weekly

OW

Output Weekly

Output Weekly

Tools, workflows and honest rankings for people who make things with AI

Tools, workflows and honest rankings for people who make things with AI

News

Rumored release dates for new AI video models (November 2026)

Ravi Menon

A close up of a movie clapper on the ground

Key takeaways

  • As of November 20, 2026, no maker has announced a dated, upcoming AI video model that we could find. Everything below is teased without a date or rumored.

  • Teased means the maker said it officially but gave no date. Kling 4.0's two-minute extension and enterprise API, Gemini Omni Pro, and MiniMax's next H-series model all sit here.

  • Rumors include Alibaba's next video model "in November," a Veo 4, a Midjourney Video V2, and a leaked Gemini Omni Pro. None has a maker-confirmed date.

  • Dates have slipped before. Public Veo 4 guesses of December 2025 and May 2026 both passed without a release, and Seedance 2.5 was reported for early July but launched July 31.

  • OpenAI has discontinued Sora, so there's no Sora successor to wait for.

Which new AI video models are coming, and when?

No upcoming video model has a confirmed release date right now. The closest things are Alibaba's reported "November" target, which hadn't shipped as of November 9, and a handful of undated promises from Kuaishou, Google, and MiniMax.

This piece sorts every claim into three labels and links the source and its date for each. It doesn't guess at dates the sources don't give, and it says plainly what nobody has said yet.

How the three labels work

Confirmed means the maker announced a specific model with a specific date. Teased means the maker said officially that something is coming but gave no date. Rumor means the claim comes from reporting, a leak, or a third-party prediction.

A rumor isn't wrong, and a tease isn't a promise. The labels only say how much weight a claim can carry. Where a report says it draws on a maker's material but the maker never put the claim on its own pages, it's treated as a rumor here.

The status table

Nothing in this table is labeled confirmed, because none of the sources gives an upcoming model with a date. Each row's source is a link, and the date in brackets is when that source was published or checked.

Model

Status

Claimed timing

Source

Kling 4.0 two-minute extension

Teased

"Coming soon," no date

Kuaishou announcement via IT Home (Sep 29, 2026)

Kling 4.0 enterprise API

Teased

Promised, no date

Magica (Oct 17, 2026)

Gemini Omni Pro

Teased

When it's "a step change above Flash," no date

TechCrunch (May 19, 2026)

MiniMax next H-series model

Teased

"Next generation," no date

MiniMax blog (Jul 31, 2026)

Alibaba next video model

Rumor

"Will also be released in November"

EachLabs (Nov 9, 2026)

Gemini Omni Pro leak

Rumor

None, a claimed internal sighting

TeamoRouter (Nov 5 leak)

Veo 4

Rumor

December 2025 or May 2026, both passed

VEED (Jul 29, 2026, edited Sep 30)

Midjourney Video V2

Rumor

None

Startup Fortune (Apr 25, 2026)

What the makers have teased

Kuaishou's Kling 4.0 launched on October 15, 2026, but two promised pieces haven't followed. Kuaishou's own announcement, as reposted by IT Home on September 29, says the extension feature that stretches clips up to two minutes is coming soon. Magica's October 17 write-up says the primary announcement describes the extension the same way, and that Kuaishou "promises an application programming interface, or API, for enterprises and developers." Neither source gives a date for either feature.

Google teased Gemini Omni Pro at I/O on May 19. Google's Brichtova told TechCrunch it will arrive when the team reaches "a step change above Flash," and Google gave no date. Google's latest is Gemini Omni 1.2 Flash, which Google made generally available on November 5 with clips up to 15 seconds. That post names no Pro model and no timeline. Google's own video documentation, checked November 20, lists Gemini Omni Flash, Genie 4, and Veo 3.1 and announces no upcoming models.

MiniMax's H3 post says "In the next generation of the H series, we plan to integrate capabilities from our M-series models." It also lists scaling the model and improving detail at higher resolutions as priorities. It's the only teased item here with no product name, and it gives no date.

What's only rumored

Alibaba's next video model is the most specific rumor. EachLabs reported on November 9 that QbitAI, citing material provided by Alibaba, wrote that a next-generation video model "will also be released in November." The article says Shanghai Securities News carried the same timing and attributed it to an unnamed person connected to Alibaba. Alibaba's own Apsara Conference summary, dated September 25, doesn't mention a new video model. EachLabs found no official use of "Wan 3.5" or "Wan 4," and no length, resolution, or audio specs have been published. Nobody has said whether the weights will be open. Wan 3.0, released in August, has closed weights, so an open release isn't something to assume.

The Gemini Omni Pro leak is thinner. A model-tracking page says an X user claimed on November 5 that he had seen Omni Pro internally, and that it mentioned 4K and 15 seconds. The page marks it "Leak, not confirmed" and says Google hasn't announced the model, its ID, its date, or its price. Treat the 4K and 15-second figures as unverified, since Omni 1.2 Flash already supports 15-second clips.

Veo 4 is a name people search for, not a product. WaveSpeed wrote on May 26 that there is "no Google announcement, product page, model card, API spec, or pricing sheet for a model called Veo 4." A May 18 TokenMix post reached the same conclusion. The rumored features of 4K, longer clips, and better consistency were never confirmed or denied.

Midjourney Video V2 rests on one secondary report. Startup Fortune said on April 25 that Video V2 is "in development" and "publicly referenced on Midjourney's development roadmap," without linking the roadmap. The roadmap claim is unverified. The report gives no date.

What's off the board or silent

OpenAI told CBS News in March, "We've decided to discontinue Sora in the consumer app and API." CBS's report was updated March 24, 2026. There's no Sora successor in the sources reviewed.

For Runway, Luma, Seedance, Hailuo, and Pika, no official announcement of a next video model turned up in the sources reviewed, beyond the MiniMax statement above. Absence of an announcement isn't evidence of anything coming or not coming, so those names stay off the table.

How often dates have slipped

Two cases show the pattern, and one shows a promise kept.

VEED's Veo 4 post said December 2025 was "the most likely release date" and offered May 2026 as the alternative. Google's I/O keynote on May 19 introduced Gemini Omni instead, and Veo 3.1 was still the current Veo model in WaveSpeed's May 26 write-up. VEED calls its dates "educated guesses."

ByteDance's Seedance 2.5 shows reported timing drifting. RuntimeWire said on June 22 that ByteDance was "pointing to an early July launch," and a BytePlus page titled its release date "coming in late July." TechNode reported the launch on July 31. The first report doesn't quote a named ByteDance spokesperson, so the "early July" figure is weaker than the others.

Kling 4.0 kept its date. The September 29 announcement said it would launch in October, and Bernama reports it went live on October 15.

Three cases aren't a rate. They show only that third-party date guesses have missed and that maker statements with a named month have held.

What nobody has said

  • No maker has given a day for Gemini Omni Pro, Kling's extension, Kling's API, or a next MiniMax model.

  • Alibaba hasn't named its November model, published its specs, or said whether it will be open or hosted.

  • Google hasn't said whether a standalone Veo 4 exists.

  • Midjourney hasn't announced a Video V2 on its public update page, as far as the reviewed sources show.

What to do with this

Plan around what's out. If you want open weights you can run at home now, the open-weight video model roundup lists what's downloadable, and Alibaba's next model isn't on it. If you're choosing a workflow, text-to-video vs. image-to-video holds up whichever model ships. For free access in the meantime, see the video generators you can use without a subscription. On the image side, the Seedream 5.5 release is a recent example of how a release shows up in docs weeks after the announcement.

Check the maker's own release page before you rely on any date here, and treat a month without a day as a target and not a schedule.

Upcoming AI video models FAQ

Is Veo 4 coming out soon?

Nobody has confirmed it. Google's developer documentation, checked November 20, lists Veo 3.1, Gemini Omni Flash, and Genie 4 and announces no upcoming models. Veo 4 dates in circulation are third-party guesses.

When is Gemini Omni Pro releasing?

Google hasn't said. It teased Omni Pro at I/O on May 19, and the only reported timing is that it will come when it's "a step change above Flash." A November 5 leak claims 4K and 15-second clips but isn't confirmed.

Is there a new Wan model due in November?

It's rumored. Reports from Alibaba's September conference say a next-generation video model "will also be released in November," but Alibaba hasn't named it or published specs. EachLabs said on November 9 that nothing had been released yet.

Will there be a new Sora?

Not from the sources reviewed. OpenAI said it was discontinuing Sora in both the consumer app and the API, and no successor has been announced.

Has Kling released everything it promised for version 4.0?

Not by the sources checked. Kling 4.0 launched on October 15, but Kuaishou described the two-minute extension as coming soon, and an enterprise API is promised without a date.