MiniMax H3 Max — Fast AI Video Drafts

Cinematic 16:9, one continuous 5-second take, no cuts. Scene: midnight creative desk — matte black wireless earbuds case on walnut desk, single hard key light from camera-left, cool monitor spill from off-frame right, faint dust motes in the beam. Camera: slow dolly-in from medium-wide to tight product hero, shallow depth of field, 35mm lens feel — one move only, no pan-orbit combo. Action: specular highlight sweeps across the case lid as camera advances; soft steam curl from a forgotten espresso cup stays blurred in deep background. Native stereo: low HVAC hum, one distant keyboard clack, no music, no voiceover. Single hero product, no readable UI text, no logos.

Vertical 9:16, one continuous 8-second take, no cuts. Scene: rain-slick neon alley at blue hour — jazz busker under a warm tungsten street lamp, wet pavement reflections, magenta and cyan sign glow in bokeh. Subject: woman in amber trench coat, acoustic guitar, eyes closed mid-performance, raindrops on strings. Camera: locked medium shot, gentle handheld micro-sway only — do not orbit. 0–3s: wide-medium, native stereo rain on awning + distant traffic wash. 3–6s: guitar plucked melody forward in stereo field, footsteps pass behind camera. 6–8s: hold framing, lamp flicker once, resolve on her smile. No on-screen text, no platform logos, no crowd cheering walla.

16:9, one continuous 8-second take from provided start frame to end frame, no cuts. Hero: frosted glass perfume bottle on black marble — gold cap closed at start, open with visible liquid glint at end; label silhouette and color unchanged throughout. Lighting: dual rim lights (warm key, cool fill), specular caustics sliding across glass during move. Camera: single slow arc orbit 15 degrees around the bottle — same height, no zoom jump. Native stereo: soft studio air tone, glass clink at 6s when cap lifts, no music. Do not invent new typography on the label, no logo morphing.

16:9, one continuous 10-second take, no cuts. Scene: high-floor corner office at sunset — floor-to-ceiling window, city haze orange, designer desk lamp pools light on a mechanical keyboard and matte notebook. 0–3s: static wide, native stereo: HVAC whisper, distant city murmur. 3–7s: slow dolly-in to medium on hands typing once, then pause; keyboard clicks sharp in stereo L/R. 7–10s: hold medium; pen tap twice on notebook, lamp filament buzz barely audible. Camera: one move only — dolly-in, no pan-orbit combo. Warm cinematic grade, shallow DOF, no on-screen text, hands only (no face required).