VideoGen says Model 1 makes 10-minute video from one prompt
The Y Combinator-backed company opened a $1 million credit pool for the model, which it projects will cost more than 50% less than comparable systems.
Key takeaways
- VideoGen unveiled Model 1 on October 20, 2026, a video model it says turns a single prompt into a finished, multi-shot video up to ten minutes long.
- VideoGen says Model 1 will cost more than 50% less than comparable models, based on its own internal validation; official pricing and third-party benchmarks are still to come.
- CEO Anton Koenig told AI Affairs that Model 1 sits at about 80% of ByteDance's Seedance 2.5 on quality, a figure with no published benchmark behind it.
- A $1 million pool of free AI credits opened to creators, filmmakers and studios, with applications at videogen.io/model-1.
VideoGen unveiled its first in-house video model on October 20, 2026, and the headline number is length. Model 1 turns a text prompt or up to 16 reference images into a finished, multi-shot video of up to ten minutes, the San Francisco company said in its launch release.
That is far beyond what most frontier models generate in one pass. ByteDance's Seedance 2.5, the model VideoGen measures itself against, produces up to 30 seconds natively. Model 1 reaches ten minutes by planning a video as a sequence of shots and carrying characters, locations and story details from each shot to the next, VideoGen says.
What it does
The model folds planning, audio and video generation, and assembly into one workflow. A "thinking harness" develops the screenplay, dialogue, shots and pacing, and the planner weighs storytelling, narrative arcs and marketing persuasion, the company says. Under the hood, Model 1 uses joint audio-video attention in latent space, decoding compressed representations into frames and sound.
VideoGen trained Model 1 across realistic film, animation, cartoons, advertising, social and training video, and says it used style-focused post-training and low-rank adapters for text-to-video and image-to-video.
The cost claim
VideoGen expects Model 1 to cost more than 50% less than comparable models, naming Seedance 2.0 and 2.5, Google's Gemini Omni Flash 1.1 and FLUX 3. That projection rests on internal validation; official pricing, official benchmarks and third-party validation are still in progress, the company said.
Quality is the softer claim. CEO Anton Koenig told AI Affairs that Model 1 sits slightly below Seedance 2.5, at close to 80% of its performance. No benchmark backs that figure yet, and AI Affairs reported it was not given access to the model and could not verify the company's claims.
The credits
Alongside the model, VideoGen opened applications for a $1 million pool of free AI credits, open to creators, filmmakers, studios, students and amateurs. Koenig told AI Affairs the credits will be given out over three to five months, with applicants judged on creativity, a viable distribution plan and a willingness to credit VideoGen. Example videos and the application form are at videogen.io/model-1.
Model 1 began a limited rollout to private partners on October 20. It is VideoGen's first model of its own; the company was founded by Koenig and David Grossman in Y Combinator's summer 2024 batch.
Applications for the $1 million credit pool are open at videogen.io/model-1; VideoGen has not published a release date or price for general access.
Sources
- globenewswire.com - the launch release: length, inputs, cost claim, credit pool
- aiaffairs.com - Koenig's 80% of Seedance 2.5 estimate, and that AI Affairs could not verify the model
- videogen.io - example videos and the credit-program application page