Ten minutes is the claim. Length is not the hard part to state
Most video models alfw.ai covers ship clips, measured in seconds. A system that returns a finished ten-minute piece, with more than one shot and a story carried across them, would be a different product category, closer to a production tool than a generator. That is the ambition in VideoGen's announcement, and the examples and credit applications are hosted on the company's Model 1 page.
Read the announcement for what it can support today, though. It contains no resolution, frame rate, generation time or price. It does not name the comparable models in the cost comparison. It does not publish a benchmark table. The whitepaper that would detail the research approach, experiments and findings is a future release. None of that makes the claim false. It makes it untested in public, which is a different and more useful thing for a reader to know before they plan work around it.
There is also a distinction the headline length can blur. Ten minutes of assembled, multi-shot video is not the same technical claim as ten minutes of continuous generation. VideoGen's own description is an assembly workflow: planning, generation and automated assembly in one system, with context carried across scenes. That is a sensible way to build long work. It also means the test that matters is not duration alone. It is whether characters, locations, sound and pacing stay consistent from the first shot to the last, and that test needs full examples watched end to end, not a launch video. The examples posted on the Model 1 page are short clips of under a minute, plus a claymation short of 4 minutes 33 seconds that the company says was generated from a single prompt with no editing. No example anywhere near ten minutes is posted.
The cost number has no denominator yet
VideoGen says its internal validation indicates that Model 1 will cost more than 50% less than comparable models. The company has not published the comparator models, evaluation settings, underlying results or final pricing. There is therefore no public basis yet for reproducing the calculation or comparing the claim fairly with Runway, Kling, Hailuo, Veo or PixVerse. The figure is a company-reported expectation, not an independently verified cost saving.
So there is currently no per-minute, per-credit or subscription figure a creator can put next to a rival and compare. When the pricing lands, the terms in this 20 October 2026 announcement should be treated as the starting point, not the settled price. If the company publishes benchmarks or pricing after launch, that will be a dated update to this story, not a quiet edit of the claim.
What is actually available today
Three things, and only three. Private partners can use the model in a limited rollout that began on 20 October, per VideoGen. Journalists can ask the company for early access and a private demonstration. Everyone else can apply for credits and watch the example videos on the Model 1 page.
The $1 million figure needs the same plain reading. It is a pool of free credits for selected creative projects, not a grant automatically available to every applicant. VideoGen says allocations will depend on project scope and scale, with selection also considering the creative work and its distribution plan. Applications opened on 20 October, and credits are to be distributed through the end of 2026. The programme offers a route to selected users, but it does not make Model 1 generally available. For a filmmaker with a project and an audience, it may still be worth applying. For a casual user, it changes nothing about access today.
If Model 1 does what is claimed, the interesting competition is not clip quality in isolation. It is whether a planning-and-assembly system can hold a story together for ten minutes at a cost creators can repeat. That question is answerable, just not from this announcement. What would make the claims testable:
- Full-length, unedited outputs that can be watched end to end.
- Resolution, frame rate and generation time.
- Published pricing, with cost comparisons against named alternatives that others can reproduce.
- Independent testing of consistency across shots and scenes.
Until some of that exists, Model 1 is best read as an announcement of intent: a system designed to assemble ten-minute videos, with its two headline numbers — length and cost — still awaiting evidence anyone outside the company can check.
Sources
- [1] VideoGen, "VideoGen Unveils Model 1, Its First AI Video Generation Model, and $1 Million in Free AI Credits", GlobeNewswire, 20 October 2026 — the company's own launch release: the ten-minute and multi-shot claims, the thinking-harness description, the internal-validation basis for the cost claim, the private partner rollout, the credit programme terms, the guardrails and watermarking, and the statement that final pricing, official benchmarks and third-party validation are underwayRead source
- [2] VideoGen, Model 1 product page — the example videos, the credit programme terms and the rollout status, checked directly on 20 October 2026Read source
- [3] Note on sourcing: the GlobeNewswire page could not be retrieved directly during preparation (it returned an access error). The release text was checked against accessible copies of the same release carried by other outlets, and the key terms were cross-checked against VideoGen's Model 1 page. All product specifications, performance expectations, cost claims and availability details in this article are attributed to the company. No independent benchmark or third-party test was available at publicationRead source