通义万相 Wan 3.0：一次生成 30 秒有声视频

Generate a 30-second, 16:9 urban chase short in a stylish motion-graphics illustration style: flat graphic shapes, refined 2.5D layering, restrained 3D volume, not live action. Image1 is the reference for the neon city, roads, and overall texture; Image2 is the only reference for the man; Image3 is the only reference for the woman. Keep both faces, hairstyles, outfits, and colors consistent. Both are martial artists fighting over one glowing ball with a white core and a purple-pink-blue halo. 0–4 s: the camera dives from the neon skyline to street level as the woman sprints into frame clutching the ball. 4–8 s: the man cuts in from a side street to intercept her; she redirects his grab and slips past. 8–24 s: the chase runs through crowds and alleys, then both grab bicycles and race through night traffic, with handheld tracking, whip pans, snap zooms, and low angles. 24–30 s: she cuts between a braking car and a bus, escapes, and glances back with a proud smile as the camera rises over the city. Only one glowing ball, always the same size.