Use the supplied little-dragon character and cinematic story keyframes as identity, world, and composition references. Create a 10-second family-fantasy trailer in three connected shots: a close emotional hesitation, a wider attempt to fly through glowing clouds, and a hopeful final reveal above the valley. Preserve the dragon’s green scales, large expressive eyes, proportions, and gentle personality. Use warm sunset light, soft volumetric clouds, restrained camera movement, natural wing effort, and a clear emotional progression from fear to courage. Add delicate wind, wing movement, one rising orchestral phrase, and no dialogue. No text, logo, character redesign, duplicate dragon, abrupt location jump, or exaggerated comedy.

Begin from the supplied living-room keyframe with the child facing a television and the same small green dragon visible inside the glowing fantasy image. Create a 10-second cinematic scene with one continuous emotional idea: golden light spills from the screen, the dragon carefully crosses the boundary into the room at toy-like scale, and the child reaches forward and gently steadies it. Preserve the child, dragon identity, room layout, television position, and warm evening color. Keep the movement slow, believable, and family-friendly; use a subtle push-in and natural eye contact. Native audio: quiet room ambience, a soft magical shimmer, tiny claws on the floor, and one warm musical resolve. No extra creature, sudden size change, distorted hands, screen explosion, text, or logo.

Create a 15-second cinematic action scene inside one modern stainless-steel elevator with two entirely fictional adult characters: one unmasked protagonist in a black leather jacket and one masked opponent in matte tactical clothing. Use exactly four readable action beats with clean coverage: block the first strike, controlled elbow counter, brief separation against opposite walls, final disarm and held standoff. Keep both bodies anatomically stable, preserve left-right screen direction, and make every contact physically motivated. Use short hard cuts between a wide two-person view, one close reaction, one side angle, and a final stable wide shot. Native audio: elevator hum, clothing movement, restrained impacts, breath, and one floor chime. No weapons entering or disappearing, duplicate limbs, teleporting, continuous morphing, gore, text, or real-person resemblance.

Prepare two short scripts around the same high-level fantasy brief, runtime, aspect ratio, visual quality target, and review criteria. Give each script its own approved opening keyframe, then animate both with Seedance 2.0 using the same motion intensity, shot count, sound requirements, and 15-second limit. Preserve the subject identity and environment supplied for each version. Judge the outputs on narrative clarity, motion stability, emotional pacing, camera readability, audiovisual rhythm, and whether the ending resolves the script. Do not change the brief after seeing the first result; the purpose is to compare scenario direction rather than optimize one side selectively.

Use the supplied creator and fictional frozen-raspberries-in-chocolate product keyframe as the visual reference. Create a 15-second vertical UGC advertisement in a bright real home kitchen. Open with the creator holding the cup near camera and saying one short natural hook; cut to a macro close-up of one frozen raspberry with visible chocolate texture; return to a genuine tasting reaction and a clean final product hold. Preserve the creator’s face, hands, clothing, cup design, product color, and kitchen continuity. Use handheld phone-camera energy without excessive shake, believable chewing, crisp wrapper and cup sounds, light room ambience, and conversational delivery. No extra fingers, melting product, unreadable invented claims, aggressive zooms, duplicated cup, or fake brand logo.

Use the supplied fictional beauty creator and unbranded lip-treatment applicator as identity and product references. Create a 15-second vertical creator video in soft window light. Keep one stable close framing: she introduces the texture in a short conversational sentence, applies one controlled stroke to the lower lip, looks into the camera, and finishes with a subtle satisfied reaction while holding the product beside her face. Preserve facial identity, lip shape, applicator size, hand anatomy, clothing, and room background. Use realistic micro-expressions, natural pauses, quiet room tone, and close clean voice. No dramatic transformation, duplicate applicator, extra fingers, smeared product, changing makeup, floating hand, readable medical claim, logo, or beauty-filter look.

Create a 15-second playful vertical commercial for a completely fictional mattress brand. A friendly original panda enters a bright minimal bedroom, performs one gentle controlled bounce on the center of the mattress, settles comfortably, and looks toward camera as the surface returns smoothly to shape. Keep the mattress geometry, bedding, room layout, panda proportions, and lighting consistent. Use three clean shots: room reveal, side view of the single bounce, and calm final product composition. Native audio: soft footsteps, one cushioned impact, fabric movement, a playful two-note motif, and no dialogue. No known mascot, repeated bouncing, broken bed, extreme deformation, duplicate panda, text, real logo, slapstick collision, or room transformation.

Use the supplied fictional fashion model walking toward camera on a clean city street as the identity, composition, and movement reference. Create a 12-second fashion campaign with a steady backward tracking shot and three clearly separated style changes triggered by natural foreground occlusions: tailored monochrome, vivid evening color, then a refined final streetwear look. Preserve the same face, hair, body proportions, walking direction, street geometry, and camera speed. Each outfit should be complete and stable before the next transition. Use polished daylight, restrained fabric motion, confident pacing, and one rhythmic electronic music bed. No body morph, face change, duplicated person, revealing wardrobe glitch, floating garments, jump in camera position, readable slogan, or real fashion logo.