Creatify Labs today launched Boreal, an AI video model that generates text-to-video and image-to-video at one cent per second and in realtime, turning out a five-second clip in about five seconds. In the benchmarks, Boreal rivals the quality of the top-tier closed models advertisers use today, at a fraction of the price and about 40 times the speed.
MOUNTAIN VIEW, Calif., Sept. 15, 2026 /PRNewswire-PRWeb/ -- A video generation API that renders in realtime at a groundbreaking rate of 1¢ per second.
Creatify Labs today launched Boreal, an AI video model that generates text-to-video and image-to-video at one cent per second and in realtime, turning out a five-second clip in about five seconds. In the benchmarks, Boreal rivals the quality of the top-tier closed models advertisers use today, at a fraction of the price and about 40 times the speed.
Two powerful factors define the new model, and they carry equal weight. The first is cost: Boreal is billed per second of finished video rather than per credit or token, so the bill for a job can be worked out before it runs. A five-second clip costs about $0.05, which is 6 to 11 times less per clip than the models in Creatify's benchmark set. The second is speed: the model serves at 1:1 with realtime, generating a second of finished video for every second of GPU time, against the minutes-long queues typical of the closed models it is measured against.
Frontier-competitive quality, verified by blind human review
Boreal is built on top of LTX-2.5, the open-source video generation model from Lightricks: a 22-billion-parameter system that generates audio and video jointly. Creatify Labs tuned that base with a single post-training recipe on advertising footage, an approach designed to turn the strongest open base model into an ad-grade generator.
To isolate what the post-training changed, every comparison in the company's report holds the two arms identical except the trained weights, using the same prompt, image, resolution and seed. In blind side-by-side review, evaluators preferred Boreal to its untouched base model in 81% of decisive comparisons (25 to 6, with 9 ties, across 40 production cases; sign test p<0.001). Running the same recipe on the previous open base moved it the same way, which the company presents as evidence that the gain comes from the method rather than a single lucky checkpoint.
Against a leading closed model run on identical inputs, Boreal matches it on the perceptual-quality measures that decide an ad and leads it on motion, at roughly 40 times the speed and 12 times cheaper at the benchmarked price. Creatify Labs is candid about the trade-off in its report: on raw fidelity the quality figures land within a few percent in both directions, on inputs where the closed model renders at higher resolution, while cost and generation time differ by two orders of magnitude. The company frames Boreal as reaching the same neighborhood on the measures that matter for advertising, holding the product on screen intact, and doing it far cheaper and faster.
The product stays the product, and the creator stays the creator
For marketers, the deciding factor is often neither price nor speed but consistency: whether the product on screen stays the exact object the advertiser shipped, and whether the person stays the same creator from the first frame to the last. General-purpose video models are prone to breaking both, swapping a label for an invented one, morphing a product into a shape that appears nowhere in the brief, or replacing the creator mid-clip. Those are the failures a viewer spots instantly, and they undo the exact thing a creator ad or a product ad is bought on.
Boreal's post-training recipe targets that fidelity directly, the half of the job Creatify calls the hardest. In the company's report, the model holds a personalized product label and on-screen name legible for a full clip, keeps a creator as one continuous person rather than swapping them mid-shot, and renders an app interface legibly where the base model paints a generic wallpaper. Its image metrics move in step: significant gains over the base model in image quality and fine-detail retention, alongside a drop in warping.
Cost and speed compound at the API level
Both gaps widen where teams generate at volume. On Boreal, a storyboard of twenty five-second shots costs about quarter of a dollar, and finishes in under two minutes. The same job on a leading closed model runs roughly $12 and over an hour of queue time, on Creatify's measurements. The pricing is designed to be legible, billed per second of video rather than per opaque credit, so an AI video generation cost can be read line by line; the throughput is designed to keep iteration immediate, so a team can regenerate a shot and see the result rather than wait out a queue.
Creatify Labs is direct about where the model stands today. Boreal is new, and it is strongest on short clips and simple scenes. The narrowest margin over its base model is on single-person talking clips, where the two are closer. The company says the same post-training recipe will follow the open frontier to newer base models as they mature, so the ad-grade layer tracks the best available open weights over time.
Availability
Boreal is available now through the Creatify Labs API, Creatify's Model Playground and AdFlow features and on fal platform. Both text-to-video and image-to-video are supported at launch.
About Creatify Labs
Creatify Labs is the research lab building lifelike AI video models, including Boreal and its avatar model Aurora. Its work centers on post-training open video base models into production-grade generators for advertising and turning them into fast, low-cost APIs. More information is available at https://labs.creatify.ai/.
Media Contact
Yinan Na, Creatify, 1 (650) 319-6694, [email protected], https://creatify.ai/
SOURCE Creatify
Share this article