MiniMaxH3.app presents itself as "H3 Studio": a single web page where you can run MiniMax H3 generations without touching an API. The interface exposes three workflows that map to the model's actual capabilities. Text to Video takes a plain prompt and produces a clip. First / Last Frame lets you upload an opening still, a closing still, or both to anchor the motion. Omni Reference accepts up to 9 images, 3 videos, and 3 audio clips (12 assets total) that you reference in the prompt as Image 1, Video 1, and so on — useful for multi-asset cinematic remixes, motion transfer, character swaps, product commercials, and scene replacement.
Output is generated at native 1440p/2K with cinematic 24 FPS motion and native stereo audio, including synced dialogue, sound effects, and ambience in the same pass. Clips run from 4 to 15 seconds, with aspect ratios spanning 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. A 768P saver tier is available for cheaper previews; any credit pack unlocks the higher resolution. The site notes that the model's safety filter rejects photos of real people, celebrities, and minors — credits are refunded but the attempt is consumed — so illustrations, AI-generated imagery, products, and scenery work best.
Important caveat: this is an independent third-party studio. The name, the showcase cases, and the model reference describe MiniMax's H3 model, but the site is not operated by MiniMax and is not an official Hailuo AI endpoint. The same credit balance also works on ByteDance's Seedance 2.5 and Kuaishou's Kling 3.0, which are offered side by side in the studio. For the current price of each pack and the exact credit cost per generation, check the on-site pricing page, since these figures change.
