武侠秋林对决

Seedance 2.0 and Minimax H3 were asked to generate a cinematic fight scene, and both models recreated the complex physics of wuxia action on their own.

The entire prompt was one line: "A cinematic wuxia sword fight between a man in black and a woman in red in a golden autumn forest".

Nobody told them how camera movement, fabric kinematics, or falling leaf dynamics work -- they just knew.

> Seedance 2.0 delivered a seamless single-take scene: the characters spin in battle, whipping up swirls of leaves, before the camera smoothly pulls up into an overhead wide shot.

> Minimax H3 took an aggressive editing approach: close-ups of clashing swords, flying sparks, a mid-air jump, and a dramatic back-to-back final pose.

The models are no longer just rendering video -> they are directing and choreographing the sequence.

Save this one -- the prompt is right there, go test your own ↓