MiniMax H3 prompts for complete audio-video direction.

Create a 10-second 16:9 product launch film for an unbranded precision desk lamp. Shot 1, 0:00–0:03: start in near darkness with a restrained lateral dolly as the lamp silhouette appears. Shot 2, 0:03–0:07: the lamp arm unfolds once with physically plausible hinges while a narrow warm beam reveals brushed aluminum texture. Shot 3, 0:07–0:10: settle into a stable three-quarter frame with the light illuminating a blank notebook. Preserve the same geometry, finish, hinge positions, desk, and camera axis. Stereo sound: quiet room tone, one mechanical click moving from center-left to center, and a soft electrical hum. No logos, floating parts, invented controls, text overlays, liquid, or unsupported product claims.

Create an 8-second 9:16 brand mood film inside an independent café before opening. Use one adult barista, one unbranded ceramic cup, and one espresso machine. Open with a close view of steam in cool window light, cut to the barista setting the cup beneath the machine, then finish as warm espresso fills the cup and the barista places it on a clean counter. Keep face, apron, cup, machine, counter geography, and lighting direction consistent. Camera is intimate and restrained: one macro insert, one shoulder-height medium frame, one final locked close-up. Stereo sound: room tone, ceramic contact, machine pressure, liquid pour. No readable trademarks, copied slogans, fake menu text, exaggerated steam, or copyrighted music.

Create a 12-second 21:9 fantasy game cinematic concept using original characters and no existing franchise designs. A lone courier enters a ruined observatory at dusk and discovers a dormant brass guardian. At 0:00–0:04, track behind the courier through the doorway. At 0:04–0:08, cut wide as the guardian's eye activates and dust falls from its shoulders. At 0:08–0:12, push into the courier's controlled reaction while the guardian remains in the background. Preserve costume, satchel, guardian geometry, room geography, screen direction, and amber-blue lighting. Stereo sound: wind through stone, boots, metal settling, one low activation tone. No gameplay HUD, logos, text, duplicated characters, weapon strike, or unsafe imitation.

Create a 15-second 16:9 original dialogue scene in a quiet late-night laundromat. Two adult friends sit opposite each other across a folding table. Use three connected shots: a wide two-shot establishing their positions, an over-shoulder shot as person A says, ‘I kept the ticket,’ and a close reaction as person B answers, ‘Then we still have a choice.’ Keep identities, wardrobe, eyelines, table objects, machine lights, and screen direction continuous. Performances are restrained, with natural pauses and small hand movement. Stereo audio: centered dialogue, machines rotating behind them, fluorescent hum, distant rain. No public-figure resemblance, subtitles, brand labels, extra people, changing props, or melodramatic camera motion.

Create a 6-second 1:1 motion-design test on a matte charcoal background. Render the exact original phrase ‘MAKE THE NEXT SHOT COUNT’ in large white uppercase geometric lettering. Reveal one line at a time through a clean horizontal mask, hold the complete phrase from 0:03–0:05, then finish with a restrained red underline moving left to right. Keep spelling, line order, kerning, alignment, and letterforms stable in every frame. Stereo sound: three dry paper-like ticks synchronized to the line reveals and one soft sweep for the underline. No extra words, logos, gradients, particles, camera shake, letter morphing, mirrored text, or decorative imagery.

Create a 10-second 9:16 ecommerce concept for a generic unbranded waterproof canvas pouch without making a performance guarantee. Show four precise beats: the closed pouch on a dry stone surface, a hand placing one folded map inside, the zipper closing once, and a light water mist landing on the exterior before the pouch is wiped dry. Use a stable overhead-to-three-quarter camera transition and neutral daylight. Preserve pouch size, olive color, zipper, stitching, hand identity, map, and surface. Stereo sound: fabric, zipper, light spray, wipe. Treat the mist as an illustrative material shot, not proof of certification. No logo, dramatic submersion, changing product geometry, impossible water behavior, extra compartments, or text claims.

Use image A only for the adult character's authorized identity, hairstyle, navy jacket, and body proportions. Use video B only for the slow semicircular camera movement; do not copy its person, setting, wardrobe, or lighting. Use audio C only for the authorized original vocal timing and words. Place the character from image A in a quiet greenhouse at dawn, standing beside a worktable and saying the supplied line once while turning toward the camera. Preserve identity and jacket throughout. Camera follows video B's motion at a restrained speed and ends in a medium close-up. Add original stereo greenhouse ambience beneath audio C. Priority: image A controls identity, this text controls environment and action, video B controls camera, audio C controls speech. No source leakage, lip mismatch, extra dialogue, plants crossing the face, or public-figure imitation.

Use authorized video A only as a motion and timing reference for one adult dancer performing a controlled turn, two steps, and a final held pose. Use image B for the new original performer's identity, silver rehearsal coat, and proportions. Replace the source setting with a plain dark studio and one overhead light. Preserve image B's identity and costume while following video A's action sequence and tempo; do not copy the source person's face, clothing, background, camera, or audio. Use a locked full-body camera with one short waist-level cut. Create original stereo sound from footsteps, fabric movement, and room reverb. No unsafe motion, duplicated limbs, costume redesign, source-background leakage, or copyrighted music.

Use image A as the exact first frame and image B as the exact last-frame composition. Create a physically plausible 10-second transition from the same empty coastal bus stop in clear late-afternoon weather to gentle rain at blue hour. Preserve shelter geometry, bench, road markings, cliff silhouette, camera position, lens, and horizon. Let clouds gather gradually, wind move grass in one direction, the first droplets appear on the glass, and practical shelter lights activate before arriving at image B. Camera remains locked. Stereo sound evolves from distant sea and wind to light rain on glass and one far-off vehicle. No people, lightning, flood, changing architecture, camera movement, text, logos, or sudden day-night morph.

Use the supplied rights-cleared character image as the only identity, age, hairstyle, wardrobe, and proportion reference. Place the adult character in a small archival reading room. Begin in a medium profile as they open a labeled but unreadable folder, move to a close-up as they notice a missing page, then hold as they say the original line, ‘Someone arrived before us.’ Preserve face, glasses, green cardigan, hands, folder, desk, lamp, and room geography. Camera: one slow push-in with a single motivated cut. Stereo audio: page movement, clock, quiet room tone, clear centered dialogue. No real-person imitation, extra text, changing identity, new accessories, dramatic score, duplicated hands, or altered folder.

Use authorized images A, B, and C as exact references for one unbranded portable speaker's front, back, controls, material, color, and proportions. Image A controls the front grille, image B controls the rear ports, and image C controls the top buttons. Create an 8-second tabletop film with a slow 20-degree dolly from front three-quarter view to rear three-quarter view, followed by one hand pressing the documented center button once. Preserve every seam, edge radius, port, button, and surface finish. Use neutral studio light and original stereo room tone with one tactile click. The images override stylistic interpretation for product geometry. No new feature, logo, floating part, liquid, changing control layout, false scale, text claim, or impossible orbit.

Use image A as the exact authorized design reference for an original stone-and-moss guardian creature. Preserve silhouette, horn count, face structure, limb proportions, stone plates, moss placement, and scale. Use video B only for the slow rising motion of a heavy body; do not copy its creature design, environment, or camera. Place the guardian in a flooded underground temple as water drains from the chamber. Stage three beats: submerged silhouette, controlled rise, and final still eye contact with the camera. Use a low wide camera that retreats slightly during the rise. Stereo audio: draining water, stone friction, falling debris, one low original breath. No existing franchise resemblance, extra limbs, changing horns, rapid attack, gameplay HUD, text, or copied source setting.