MiniMax H3 AI Video Generator Video

• Multimodal understanding: a prompt, photo, clip and audio file are read together as one brief rather than four separate settings.
• Omni Reference: combine up to 12 files — 9 images, 3 videos, 3 audio — to steer identity, motion, framing, edit rhythm, atmosphere and sound.
• Reference a character, a move, a camera or a voice: point at a person for the look, at footage for the performance or camera move, at a recording for the voice.
• Precise video and audio editing: swap a person or object, change the background, light or dialogue line while the rest of the shot stays put.
• Up to 2K in six aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16, chosen as a setting rather than cropped afterwards.

The page lists trailers and brand films with multi-shot cuts up to 15 seconds and the score already mixed in; ads and e-commerce, where one product photo becomes a 2K ad with logo and text intact; and UI and product motion for landing pages, app screens and posters.

Other listed uses include cinematic stories that keep the same faces from shot to shot, game visuals that keep a character while borrowing a move and camera, and social clips in 9:16 with stereo sound. The page also names social content creators, advertisers, e-commerce sellers, marketing teams, agencies and studios, indie filmmakers, game creators, educators and product teams as intended users.

The page shows a promotional offer of 14,400 credits for $23 instead of $468, described as 50% off, with a prompt to claim the credits. It states that subscribing, upgrading or switching to an annual plan saves up to 50%.

Free credits are said to arrive at sign-up with no card needed, and there is a "Start for free" call to action. The page also mentions that you can start with MiniMax H3 and switch anytime to other models such as Seedance 2.5, Veo 3.1, Wan 3.0 and Sora.