Boreal
Text to video and image to video at 1 cent per second.
See Boreal in action.
Video by Flycatcher Studio — flycatcherstudio.com
Efficient by design.
A fraction of the cost, generated in seconds. Built to run at volume, so you can iterate freely.
per second of video
cheaper than competitors
to generate a 5s clip
Generate every version.
Keep the best.
Every video model gets things wrong sometimes, so people regenerate constantly. Boreal makes that cheap: run several versions at once, keep the one that works, and still pay less than one generation elsewhere.
"A paper cup drifts across the pond, turning slightly, a duckling sitting still inside with a straw trailing in the water. He tilts his head to watch the reeds pass. Camera low at water level, drifting alongside. Audio: soft water, reeds, distant ducks."
A fraction of the price.
/ second
Around $0.05 for a 5-second clip.
Against the models in our benchmark that's 10x to 47x less per clip.
Billed per second of video, not per credit.
Cost for a ~5-second video, the average clip length in our study. Aurora bills per second: $0.07 at 480p, $0.14 at 720p.
Approximate list prices for a ~5s 720p clip on fal.
AI video generation cost, priced per second so you can work out the bill before you run it.
Made for everyone
01/03
Marketing
Ad creative and UGC in volume. Generate a dozen versions of every concept and keep the winners.
02/03
Product
Hero shots, demos, and lifestyle scenes from a single product photo, with the product kept intact.
03/03
Social
Short-form in every direction: characters, trends, and b-roll, cheap enough to post daily.
Built on open weights.
Boreal is post-trained on an open base model — 22B parameters, audio and video generated jointly. We ran one post-training recipe on ad footage and kept everything else identical.
In blind side-by-side review, people preferred Boreal to its untouched base in 81% of decisive comparisons (n=40, p<0.001). Only Seedance 2.0 scores higher on raw quality — at roughly 30× the cost per second and 40× the wait. Dollar for dollar, nothing gets closer to a real ad.
It's a new model and we'll say where it's weakest. Short clips and simple scenes are where it's strongest. Single-person talking clips are where the gap to the base narrows most, at 60%.
Launch benchmark · 20 ad prompts
Best ad quality per dollar
Pass = the model beats or ties the human-made reference (n = 20). Cost is per second of finished video. Kling's score is calibrated from its direct match-ups vs Boreal and Seedance.
