A Chinese AI startup just made it cheaper to generate professional-quality video than ordering lunch in most major cities. MiniMax Group released its H3 multimodal video generation model on July 31, capable of producing 2K resolution videos with stereo sound at what the company claims is less than one-third the cost per second of mainstream competitors.
What MiniMax actually built
The H3 model generates videos up to 15 seconds long at native 2K resolution. The model accepts text, images, videos, and audio simultaneously, meaning you can feed it a brand logo, a script, a reference clip, and a soundtrack and get back something coherent.
The practical applications lean heavily commercial: advertising, branding, e-commerce product videos, and gaming assets. MiniMax is positioning H3 not as a toy for social media creators but as a production tool for businesses that currently spend thousands on video content.
On pricing, the company claims H3 produces 2K video at under one-third the cost per second of mainstream models. For lower-resolution 768p output, the cost drops to under half of what competitors charge for 720p equivalents.
