1. Segment Definition

Cinematic realistic texture. Utilizing a 'reference-image-driven living world system,' blending poetic Xianxia tension, space-misleading comedy, restrained performance, chilly silver-blue morning, warm sunrise reflections, natural volumetric mountain mist, and realistic wet stone and water physical effects. The final reversal must be achieved entirely through visibility changes. Geographic structure must be continuous; no sudden scene changes or terrain deformations allowed.

2. Reference Responsibilities
Seedance 2.0 supports image, audio, and video references; thus, all reference images are treated as a single world setting library, not isolated backgrounds. Reference images collectively determine:
Continuous terrain, architectural logic, spatial scale, material era, vegetation ecology, water levels and systems, weather status, cloud movement, mist density, main light direction, reflection relationships, sound distance perception, and realistic character movement paths.

Pre-generation Internal Environment Deduction:
Execute a silent environmental deduction first. Reorganize a unique new space based on all reference images. This space must be physically unified, structurally continuous, realistically schedulable by the camera, have stable foreground/midground/background anchors, and maintain the environment DNA of the same world.

3. Character Lock
Character ID A | White-clad Sword Immortal: Same 25–30 year old East Asian female sword immortal. Strictly maintain: @Image 1 (same face), black long hair, slender body proportions, white embroidered silk Hanfu, jade accessories, white cloth boots, and a unique long silver sword.
Character ID B | Green-clad Rival: Same 20–25 year old East Asian female rival. Strictly maintain: @Image 2 (same face), black braided hair, petite body proportions, green Hanfu, dark cloth boots, and a unique dark steel sword.

4. Storyboard Design
Shot 1 | 0–5s | Wide or Extreme Wide Shot
Goal: Establish a misleading spatial relationship, making the audience and characters believe there is a bottomless canyon ahead.
Content: Two characters stand fully from head to toe on two ancient stone platforms covered in wet moss. Heavy mist blocks the view between them, making it look like a bottomless chasm.
Anchors: Waterfall disappears into the mist; pine branches sway naturally at different depths; water droplets flow slowly on stone; distant tiled buildings appear and disappear in the mist; rising sun pushes the mist horizontally.
Performance: Both remain cautious and restrained, as if facing a great enemy. Establish the spatial misdirection first.

Shot 2 | 5–10s | Medium or Cowboy Shot
Goal: Both misjudge the spatial scale and prepare high-consumption sword-crossing moves.
Continuity: Maintain same characters, outfits, swords, geography, mist/platform logic, and environment anchors (pines, waterfall, tiled buildings).
Action: Both enter a charging state. Wide sleeves are lifted by airflow; sword blades emit low-frequency resonance; mist swirls around ankles. Dignified, formal, and restrained posture.
Dialogue: White-clad immortal says calmly: 'This sword strike crosses the canyon.' Green-clad woman replies: 'I'll be waiting.'
Camera: Slow orbital shot around the duo. Anchors must remain visible to verify space.

Shot 3 | 10–15s | Close-up or Extreme Close-up
Goal: Complete the reversal through visibility changes. The truth must come from clearing the existing environment, not generating a new one.
Event: A natural mountain wind clears the last layer of mist. Reveal: The 'bottomless canyon' is actually just a three-step-wide ancient mountain irrigation canal. Dark water and vertical stone reflections caused the depth illusion.
Hard Constraints: Continuity of stone edges, waterfall entry, moss, pines, tiled buildings, sun direction, and mist logic. Nothing should morph suddenly.
Performance: Both freeze with swords raised. Green-clad woman looks down slowly and says: 'Maybe... just walk over?' White-clad immortal silently sheathes her sword, steps across on three existing rocks, and says impassively as she passes: 'Saves spiritual power.'
Ending: Extreme close-up of the green-clad woman trying to suppress a smile. The formerly grand 'canyon waterfall sound' now sounds like a small bubbling canal.

5. Audio-Visual & Physical Requirements
Native synchronous Mandarin dialogue; consistent spatial sound; unified physical logic for wind, mist, water, and reflections; continuous natural motion and parallax; realistic movement for silk, hair, and droplets.

6. Technical Specifications
Strict duration: 15 seconds; 16:9 aspect ratio; 3 continuous clear shots; only 2 visible characters; no subtitles.

7. Negative
blurry, bad quality, low quality, low resolution, noisy, jpeg artifacts, watermark, text, error; deformed, mutated, bad anatomy, poorly drawn hands, bad composition, out of frame, disfigured; inconsistent character, changing clothes, face morphing, background shift, glitching cuts, disappearing props