Lovart adds MiniMax H3 for cheaper 15-second cinematic video
By Steven Van ·
The model renders up to 15 seconds with multimodal references and consistent characters, at less than half the usual cost.
Lovart has added MiniMax H3, calling it the world's #2 video model, and the pitch centers on a specific tradeoff most creators have had to accept until now: production-quality AI video at less than half the usual cost.
15-second cinematic output
On Lovart, H3 renders up to 15 seconds of cinematic output with fluid, high-fidelity motion, longer than most video models default to and long enough to carry an actual scene rather than a single beat. It accepts multimodal references across text, image, video, and audio simultaneously, and keeps characters, products, scenes, and styles consistent across a generation, the piece that usually breaks down when a shot needs to match an earlier one.
No more choosing between quality and cost
Lovart's framing calls out ads, music videos, and TV commercials specifically, the kind of work where quality has traditionally forced a higher price tag. MiniMax has built H3's whole go-to-market around undercutting that assumption, positioning it against pricier models like Seedance while claiming comparable output.
One of six simultaneous launches
H3 landed on Lovart the same day it appeared on Krea, Magnific, Leonardo, Pika, and Runway, a coordinated rollout that puts the same model in front of nearly every major AI creative platform's user base at once.
Sources: Lovart on X.