Wan 2.6: 영화적 일관성과 완벽한 립싱크를 갖춘 15초 AI 비디오.
Alibaba의 차세대 비디오 모델—더 스마트한 프롬프트, 향상된 오디오 동기화, 그리고 비교할 수 없는 캐릭터 일관성.
체험해 보기
핵심 기능
다중 샷 내러티브 생성
대부분의 오픈소스 비디오 모델은 단일 연속 클립을 생성하며, 종종 구조나 일관성이 부족합니다. WAN 2.6은 간단한 프롬프트에서 직접 다중 샷 내러티브를 생성하는 기능으로 주요 돌파구를 도입했습니다.
Prompt
The scene unfolds in first-person POV inside a bright, refined modern kitchen. Natural daylight pours across walnut flooring and matte gray cabinetry, giving the space a calm and polished atmosphere. The viewer takes three to four slow, steady steps forward while holding an empty celadon-green porcelain bowl with both black-gloved hands. Ahead stands a built-in double-door refrigerator. The left door features a softly glowing dispenser slot, with faint vapor curling from its edges. When the viewer reaches the refrigerator and lifts the bowl beneath the outlet, a gentle mechanical hum begins. From the small dispenser opening, the plating sequence unfolds with precise, almost ritualistic elegance. First, a smooth stream of deep orange lobster bisque flows into the bowl, circling and rippling as it settles. Moments later, tender pieces of lobster claw and tail meat descend into the center, their pink-red surfaces glistening in the hot broth. A thin ribbon of cream follows, tracing a delicate spiral across the bisque. Finally, micro herbs and tiny gold flakes drift down, completing the dish with a soft visual flourish. The celadon glaze of the bowl reflects the bright natural light, while the warm tones of the bisque shimmer gently on the surface. Subtle sounds fill the space: soft footsteps on the wooden floor, the quiet friction of gloves against the bowl, the rising hum of the refrigerator, the thick pour of bisque hitting the ceramic, the gentle plop of lobster pieces, the light drizzle of cream, and the faint sprinkle of herbs and flakes. Altogether, the moment blends mechanical precision with the warmth and intimacy of fine dining, presented through the calm rhythm of first-person ASMR realism.
Final outcome
참조 기반 비디오 생성
WAN 2.6은 비디오 참조 생성을 지원하여 사용자가 입력 비디오로 모델을 안내할 수 있습니다.
Prompt
character1 is eating dinner with character2 in a restaurant
Final outcome
15초 긴 비디오 생성
많은 오픈소스 모델은 매우 짧은 비디오(일반적으로 2-5초만) 생성에 제한되어 내러티브 깊이를 제한합니다. WAN 2.6은 최대 15초 길이의 비디오를 지원하여 이 장벽을 깨뜨립니다.
Prompt
Generate an approximately 15-second cohesive narrative video. Story: A medieval knight awakens on a storm-swept meadow after a fierce battle. First 5 seconds: A slow circling shot reveals his mud-covered armor, scattered debris, and lingering flashes of lightning in the dark sky. Middle 5 seconds: The knight rises, grasping a sword embedded in the ground. The camera pulls upward from a low angle, emphasizing the determination in his eyes. Final 5 seconds: He begins running toward a distant ruined castle wall as the camera follows in a handheld-style tracking motion, tall grass brushing past the lens to create dynamic depth of field. Maintain scene continuity, natural body motion, and cinematic epic atmosphere throughout.
Final outcome
