(Not a complete prompt; feel free to add)
Images 1–4 are consecutive keyframes, simulating an old telescope searching for the MINIMAX device. Start with a blurred handheld shake, then quickly pull into rack focus on image 1. Between images, a telescope-scan transition (whip pan, motion blur, optical trailing, exposure flicker) switches at the most blurred moment and quickly stabilizes back into focus. Throughout: fixed dual-circular lens mask (black feathered vignette, position/size/edge blur absolutely consistent, no deformation or drift) — only the interior moves. In image 2, fabric sways gently in the wind with the MINIMAX text riding the fabric's undulation while staying readable; in image 3, the figure is like a casually captured fashion passerby, naturally turning and swinging arms; in image 4, the figure adjusts glasses or lifts the chin slightly, projecting the cool, relaxed attitude of a fashion campaign. Red text loads subtly with the focus change, appearing slightly soft at low opacity, fading in and sharpening within 0.3–0.5s (a slight vertical slide or letter-spacing breath is fine), fading out or being swept away by motion blur before the switch. No rotation, bouncing, or large fly-ins/outs. Wes Anderson-style voyeuristic 35mm film texture (slight grain, soft highlight halation, restrained color, red accents), minimalist and premium with a touch of playful wit. Absolutely no added people, vehicles, buildings, or logos; keep the original core composition and the MINIMAX device unchanged.

Epic space-opera theatrical short trailer
(Not a complete prompt; feel free to add)
Faster rhythm, grand but not draggy. Quick hard cuts, bridge vibration, intense light flash, short blackouts, warp impact. Text is movie-trailer-style wide-tracking title packaging; the font shouldn't be pure white — restrained glow and material texture, faint edge halation. Text motion includes fading in from dark deep space, being swept bright by starlight, letter-spacing expansion, afterimages, faint glow, and blackout flashes. (Not a complete prompt; feel free to add details)

Sci-fi suspense film short trailer
(Not a complete prompt; feel free to add)
Realistic cinematic, high-contrast lighting, tight rhythm. Image 1 serves as the overall atmosphere and style reference. Image 2 serves as the protagonist character reference.

Shot 1 — ultra-wide establishing shot. A giant circular cosmic gate almost fills the frame; the character is only a small back-figure at the bottom, slightly right of-center. The ground is wet and reflective; the gate's center is pure dark. The camera slowly pushes forward. The main title fades in from the dark edge, blurry then sharp: "THE STARS WERE LISTENING" — extremely narrow, heavy, all caps, dark red mixed with rust red, with slight grain and fogged edges. Audio: deep low-frequency pulse, faint metal vibration in the distance, a soft hit when the title sharpens. → Hard cut. (Not a complete prompt; feel free to add details)

(Not a complete prompt; feel free to add)
Create a 16:9 horizontal premium fashion brand video. Overall mood, scene, and film texture reference image 1; bag asset references image 3; character asset references image 2; brand ending logo references image 4. This is a fashion campaign selling clothes and bags — the overall tone should be high-end, cool, and restrained, but the edit can't be boring: it should be more dynamic, more fashion-rhythmic. Don't make a regular drama or an e-commerce ad. The core story is still simple: by a desert highway and a vintage car, a woman returns to the trunk, opens it, takes out a black bag, shares a brief quiet beat with the man standing by the car, then walks away with the bag. The bag and clothing should flow naturally into the character's movements and become part of her presence.

(Not a complete prompt; feel free to add)
Image 1 is the overall texture and atmosphere reference; image 2 is the character's physical appearance reference. Generate a 15-second, 16:9 horizontal trendy fashion short. Keep the character consistent: platinum-blonde long hair, narrow black vintage sunglasses, high-gloss black patent trench coat, cool confident fashion expression, orange reflections of firelight on the patent leather. Overall style is a film-grain fast-cut fashion ad blending a night fire scene, black smoke, orange-red flames, VHS glitches, CCTV broadcast interruptions, 90s analog film grain, scan lines, chromatic aberration, light leaks, flash-white transitions, and slight frame shake.

(Not a complete prompt; feel free to add)
15 seconds, 16:9 horizontal video. A hybrid image blending live-action footage of an evening kitchen with hand-drawn glowing animation. Sunset warmth still lingers by the window; the lived-in kitchen has an old wooden table, a half-washed mug, a slightly fogged glass bottle, and a hanging dishcloth. The frame carries the slight handheld shake of a phone shot one-handed, the hesitation of close focus, exposure fluctuations from the backlight, and rough noise in the shadows. Don't let it look like a tidy ad image — give it the natural texture of someone hastily filming an inexplicable event at home. No giant eyes, split mouths, fangs, intimidation, lunging, sudden black screens, or jump scares. Sound uses only the kitchen's ambient sounds, the dishcloth's friction, mugs lightly clinking, the tap dripping, the shooter's footsteps and faint breathing, plus the hand-drawn creature's soft electronic tone and small calls.

(Not a complete prompt; feel free to add)
15 seconds, 16:9 horizontal short. Live-action footage of a late-night self-service laundromat fused with hand-drawn glowing animation. A small laundromat with slightly flickering fluorescent lights, running washers, plastic laundry baskets, an old bench, a stray sock on the floor — quiet, with a faint nostalgic mood. One-handed phone shooting with clear shake; the white fluorescents make exposure flicker; glass surfaces carry environmental reflections; focus lags when the lens moves close to objects. Don't make it polished like a commercial ad — the texture should feel like real documentary footage, someone wandering into a laundromat late at night and chasing a strange vision with their phone.

(Not a complete prompt; feel free to add)
Style: dark-pop / cyber-grunge / rap music video, realistic high-couture texture, film-magazine texture, high contrast but not cheap. Overall reference: late-90s to early-00s independent magazines, photocopy paper, film scans, underground music posters, and zine collage aesthetics. The frame has coarse grain, slight film jitter, halftone dots, print fuzz, and scan-line misregistration. The edit is fast and uses only hard cuts — no fades or soft transitions. Text packaging style and texture reference the image.

Remove the green-screen background from video 1 and replace it with a fairy-tale background similar to video 2. Background elements must fully match the character's actions in video 1. Adjust the lighting on the character in video 1 so it completely matches the new background.

(Not a complete prompt; feel free to add)
Create an animated-poster video. Keep the original image's gallery white frame, inner frame, red-white-black palette, 3D figurine feel, and layout structure unchanged. When the text appears, pair it with a lively text-entry sound effect.

Hand-drawn 2.0
(Not a complete prompt; feel free to add details)
Overall video: 16:9 horizontal, about 10–15 seconds total, 4 shots, roughly 3 seconds each, connected by brisk hard cuts. The picture is live-action smartphone footage of a "seaside market stall": light-colored stalls, awnings, shell ornaments, a cold-drink stand, wind-blown cloth curtains, distant beach and pale blue sea, bright, light, summery. The shooter's right hand holds a thin blue-green pen in a normal, natural, correct writing grip and enters from the bottom of the frame. Each shot points at exactly one real object; the object pointed at completely transforms into rough hand-drawn 2D animation, while all other background and non-pointed elements stay live-action.

AR / augmented-reality creative
The protagonist appears inside a white rectangular frame in the scene. He grips the frame's edges with both hands, first peeking out head and upper body, then planting one foot on the white line below and crouching to build power. The camera pushes in slightly; the protagonist then springs out of frame with the built-up force.
Cut to the space of IMG1. The protagonist lands on top of the white 3D text in the scene, sits on the top edge of the letters with legs dangling. After a brief pause, he turns to look at the background figure on the right side of the frame — a clear moment of eye contact.
Then the protagonist tosses a thin white elastic rope toward the upper-right corner, using it to swing in an arc. The camera follows with a fast lateral move, keeping the flat-poster spatial gliding feel.
Cut to the space of IMG2. The protagonist descends along the arc and steps onto the play button. He taps the button with his foot or hand, triggering a clean press feedback.
Finally cut to IMG3.png. The protagonist runs from the play button to the right side of the frame, returning to his original spot on the background card and standing still. Throughout: fast lateral tracking, beat-synced quick cuts, slight push/pull, and clean flat-space transitions. Keep white, gray, and the original design elements. No blood, fighting, weapons, superhero logos, or copyrighted characters.

(Not a complete prompt; feel free to add details)
The short's precise opening frame. Keep the electric-blue background, the white "MiniMax" on the left, the light-blue octopus logo at center, and the white "H3" on the right fully intact. Image 2: the target waveform logo reference after the octopus logo's linear morph — also the core graphic running through the second half. Strictly maintain the waveform logo's path structure, number of curves, rounded endpoints, and overall proportions. Generate a strictly 15-second, 16:9, 4K, 60fps MiniMax H3 abstract tech MG brand film. The film expresses H3's design philosophy of moving from specialized tasks to a general multimodal model: language, image, video, and audio are no longer isolated from each other but connected by the same context, the same architecture, and the same continuous line. Overall: premium, clean, nimble, artistic. Large areas of electric blue, deep black-blue, light cyan, and a few purple color blocks, paired with clean bold white text. Generous negative space; no dense icons or stacked information. The film's core moving element is a single light-blue continuous line with a white glowing node at its leading end.

(Not a complete prompt; feel free to add details)
Create a 15-second premium motion-graphics brand film for the AI brand "MiniMax H3."

【Text that must appear accurately】Brand name: MiniMax H3
Brand slogan: Intelligence With Everyone
No other readable text may appear in the frame besides these two strings.

【Core creative concept】"One becomes everyone."
Intelligence starts from a single tiny node and, through connection, sharing, and collective growth, becomes an open intelligent network belonging to everyone.
Build an original brand visual system from three basic elements:
a tiny 2D solid dot
the connecting bridge represented by the letter "H"
the three layers of intelligence track represented by the number "3"
All dots, lines, tracks, glyphs, and transitions must evolve continuously from this base element set. Every change has a clear visual cause-and-effect, forming a unified brand visual DNA.

All-Chinese text animation
(Not a complete prompt; feel free to add details)
An experimental Chinese dynamic-typography film on a pure black background. Only white Chinese text is allowed in the frame, with a very small amount of dark-red text for misalignment or emphasis. No people, landscapes, photos, illustrations, icons, products, logos, English titles, fixed UI, numbering, timers, borders, grids, or decorative graphics of any kind. This is not a subtitle video; it's a high-density kinetic-typography short with "the Chinese characters themselves" as the sole protagonist. Text appears, enlarges, rotates, compresses, collides, breaks apart, multiplies, inverts, and reassembles at high speed on the black plane. Overall like a graphic designer's Chinese kinetic-typography experiment, a type-motion showreel, or a music-event opening visual. Dark, cool, sharp, rational, inorganic, with a slight edge — yet premium, restrained, and unified.

TouchDesigner-style motion
(Not a complete prompt; feel free to add details)
@video1 is the only base asset. Strictly keep the original video's character, facial features, clothing, action, scene, overall color, camera movement, duration, and frame rate completely unchanged. Only add TouchDesigner-style motion-tracking graphics on top of the original footage.
1. The color fill of every rectangle must be 0% transparent.
2. No red, cyan, or any colored mask may appear inside the frames.
3. Inside the frames, only brightness, grayscale, exposure, and light/shadow relationships from the same position of the original video may be processed.
4. All frames must be complete, simple, continuous closed rectangles in either red or cyan, with no decoration or nodes at the four corners.
5. A frame may not fully contain another frame.

Not a complete prompt; feel free to add
A period wuxia suspense short-drama scene, 16:9 horizontal cinematic frame, night bamboo forest, cool blue, ink green, grey-black low-saturation tones, thin mist, fine snow drifting through the air. Overall: cold, austere, restrained, with the tension of a jianghu sect investigating and handing over classified intel. Real period-drama texture, shallow depth of field, cinematic lighting. No modern elements, no subtitles, no text, no watermarks. Scene: deep in a night bamboo forest, dense vertical stalks behind, cold white fog light seeping from the distance. Blurred bamboo leaves in the foreground create a "peering through the grove" spatial layering. Cold white side-front soft light illuminates the face; backlight from behind creates a dark-front/bright-back cinematic look; shallow depth of field turns bamboo leaves, snowflakes, and the background grove into soft bokeh. Visual style: period wuxia cinematic, cold-blue ink-green snowy bamboo grove, shallow DoF, foreground bamboo occlusion, face close-ups as the primary framing, shot-reverse-shot editing, restrained but tense rhythm. No subtitles, no platform watermarks, no modern clothing, no modern buildings, no animation feel, no excessive skin-smoothing, no bright daylight, no comedic performance.

Not a complete prompt; feel free to add
An indoor family-argument short-drama scene. Vertical-screen short-drama texture, real human performance, Chinese family/small-restaurant interior, warm lighting, red decorations and calligraphy scrolls in the background, shallow DoF, intense emotion, tight editing. Performance: real short-drama acting, no exaggerated stage delivery. Qin Haoxuan's tone is angry, aggrieved, urgent rebuttal; the middle-aged woman's tone is sharp, dominant, aggressive interrogation. Strong confrontation between the two, escalating step by step. Visual style: 9:16 vertical, phone short-drama texture, live-action feel, shallow DoF, warm indoor light, mostly medium close-ups, frequent shot-reverse-shot editing, lived-in background. No sci-fi, no period drama, no animation feel. No subtitles, text, platform watermarks, or overlays of any kind in the frame.

Not a complete prompt; feel free to add
Generate a 15-second, 9:16 vertical overseas live-action vampire romance short-drama trailer. Male and female lead appearance reference image 1; scene reference image 2. Keep both characters consistent, real-human texture, premium short-drama quality. Story: an innocent human woman wanders into a forbidden castle zone and accidentally wakes the sleeping vampire nobleman. He senses something about her connected to an ancient war and becomes intensely possessive and dangerously interested. She fears him but doesn't fully submit, resisting his pressure. Overall style: overseas ReelShort / DramaBox vampire romance trailer feel. Darkly romantic, dangerous attraction, fate, heavy control, oppressive gloom, high-contrast reversals. Premium, restrained, tight — like the first 15 seconds of a hit short drama. No gore, no cheap horror, no Halloween feel, no modern street feel. Frame: 9:16 vertical, built for TikTok / ReelShort / DramaBox. Characters mostly in medium close-ups, close-ups, and extreme close-ups; in vertical format, emphasize face, eyes, sense of pressure, and relational tension.

(Not a complete prompt; feel free to add)
Generate a vertical 9:16 premium fashion eyewear commercial. Overall reference the provided video's storyboard rhythm, cutting speed, white-studio texture, and cool fashion mood. The image is a minimalist white studio, seamless white background, strong couture-ad feel — clean, simple, sharp, avant-garde, top-tier international fashion editorial quality. Main visual character references image 1: two full-body female models — one Black, one white — maintaining their premium clothing, body posture, white-studio lighting, runway presence, and overall cool attitude. Both wear futuristic premium eyewear; the frame design references image 3, emphasizing wraparound curved surfaces, sharp geometric cat-eye/goggle hybrid silhouettes, mirror reflections, streamlined temples, and premium fashion-accessory texture. The two models' facial details reference image 2.

Product feature understanding & visual presentation
(Not a complete prompt; feel free to add)
A 360° showcase of a premium office Herman Miller black ergonomic chair. Close-up of the breathable mesh back with airflow visualization. Lumbar support + ergonomic curve engineering animation. Multi-dimensional adjustment demo of armrests/seat height. Long-hours office scenarios for designers/programmers/creatives, with easy, natural postures. A 3D skeletal support diagram for "sitting without fatigue." Spatial aesthetics of interior display. At the end, "Inspiration and comfort, online at the same time" fades in. High-end minimal, cool tones, professional futuristic, slow camera rhythm. Product details reference image 1; product image reference image 2.

(Not a complete prompt; feel free to add)
A website page, website UI design, website motion. The video demonstrates a smooth web-page scrolling-down effect. A product landing-page UI/UX demo video with explosive, dynamic energy; the core display subject is product image 1. The page uses bold, slanted, oversized sans-serif typography for loud layouts. Background: high-speed dynamic lighting and shadow, dark carbon-fiber or breathable-mesh textures interweaving and transforming. The video shows a tightly paced, powerful page-scroll-down effect, plus strong UI interactions on mouse hover like visual enlargement and color inversion.

(Not a complete prompt; feel free to add)
Simulated website UI design: first the title at the top slides down into view, then the text bar at the bottom slides up, and the car's lights change from dim to red.

(Not a complete prompt; feel free to add details)
Use Image 1 as the precise UI/layout/interface style reference.
The mouse cursor in Image 1 is also the precise cursor style reference: the cursor must keep Image 1's size, shape, proportions, semi-transparent texture, soft glowing edge, and refined premium feel. Do not generate an oversized system-default cursor, a crude black-and-white arrow, or multiple cursors.
Use Images 2–5 as the precise companion-creature references, mapped as follows:
Image 2 = card A = MIRRA POD
Image 3 = card E = LANTERN
Image 4 = card F = MOSS MITE
Image 5 = card C = VELI KIT
Generate a 15-second, 16:9 horizontal video.
The camera is completely fixed — no push, pull, pan, tilt, or cutting.
Keep Image 1's interface structure, typography, panels, font hierarchy, icons, corner radii, card grid, and overall composition stable, elegant, and clearly readable.
The interface mood is premium, minimal, dreamlike digital companion-creature codex.
No extraneous text, no garbled characters, no UI distortion, no interface drift.
Overall style: dream-fog, soft focus, micro-fuzz, translucency, gentle glow, refined minimalism.
All four companions must be very cute, nimble, soft, and spirited.
Motions must be clean, refined, and interactive.
The interface itself stays stable; only the cursor, selection states, main title, the currently featured companion, and a small amount of skill effects animate.

Fixed camera. The robotic arm descends, grabs the red cube, moves forward a distance, then sets it down.

Precise background replacement
Change-the-background prompt: replace the desk in the original video with a standard office workstation desk, and replace the overall background with an office environment featuring blinds and metal filing cabinets. The new desk's perspective must match the original lens exactly; the desktop material must naturally inherit the original video's light and shadow projection; edge shadows and highlight direction must strictly match the original scene's light source. The background blinds and filing cabinets' depth-of-field blur must align with the original video's focus, preserving the original sense of spatial depth. Throughout this process, the original video's camera trajectory, the robotic arm's complete motion sequence and speed, the relative positions and occlusion relationships of other objects, and the overall lighting atmosphere must remain 100% unchanged, ensuring the replaced elements blend seamlessly with the original frame.

(Not a complete prompt; feel free to add)
Claymation style. A fox sprinting at full speed reaches the cliff edge, leaps without hesitation, and performs a tension-filled, slow-motion heroic jump over a vast lava canyon. The camera whips beneath the fox's belly mid-air (a high-dynamic camera move), fully revealing the terrifying depth of the abyss and the extreme stretched-motion moment of the fox's clay body in the air.

(Not a complete prompt; feel free to add)
Use image 2 as the fixed character reference, keeping the black half-tied long hair, silver openwork hair crown, indigo-blue hair ribbon, light layered Hanfu, translucent blue outer robe, deep-blue waist sash, silver floral clasp, and long tassels consistent. Use image 1 as the storyboard camera and rhythm reference. The picture is high-quality 4K 16:9 Chinese-style 3D, cinematic xianxia quality, passionate, solemn, and fate-driven. The shots proceed through the storyboard's content in order, each transition natural — not like a PPT. Character faces may be shown only in close-ups or extreme close-ups; long shots must be backs, side-backs, or empty environmental shots — no full-face long shots.

(Not a complete prompt; feel free to add)
Generate an otome-game male-lead character PV. Character reference: use image 2 as the strict character-identity reference. Keep the same character's face, hairstyle, body proportions, costume design, material detail, and otome-game CG quality consistent.

(Not a complete prompt; feel free to add)
Camera: first-person · eye level · handheld game footage. Picture: the frame simulates a player operating a modern-warfare FPS — two hands holding an assault rifle, advancing slowly around the perimeter of a military base. The player moves forward along a road beside cover; the crosshair sweeps the corridor ahead, pauses briefly, fires a few shots at a distant target point, then continues advancing — like real player footage. Light: the cool natural light of a modern military base mixes with smoke and firelight; the image is realistic and sharp, with 3A-game texture on metal weapons and battlefield dust. Camera: slight handheld shake follows the player's movement — slow forward first, then small left/right observation, slight recoil shake when firing, and finally a steady advance forward.

Human-computer interactive game
(Not a complete prompt; feel free to add)
Use the first image as the first frame and the second image as the strict last frame to generate an otome-game visual-novel interface transition video. Overall feel: a high-end Chinese otome-game romance interactive interface, an ambiguous moment before and behind the stage — a natural transition from "choosing to watch his performance" to "Han Xu being drawn by the heroine's words and responding with an intrigued reaction." UI text, options, and dialogue boxes appear in refined otome-game style; the transition is silky; the mood is ambiguous yet restrained.

Scene-based dramatic performance
(Not a complete prompt; feel free to add)
The protagonist is image 1. Cinematic, realistic style, live-action realism. Front-facing near-shot camera; the frame focuses only on the male and female leads' upper bodies. The woman's face barely appears in the video; she wears a red dress that bares her waist and back, with only occasional neck, shoulder, arm, waistline, hem, and hand details visible. The man wears a flawlessly tailored black bespoke shirt. The scene is the red-black-toned interior from image 2 — dim overall, low saturation, luxurious, restrained; the background is slightly blurred; a red velvet sofa, a black-and-gold marble side table, and warm low light sources together build a mature, ambiguous date atmosphere. The action happens in front of the red velvet sofa from image 2. The man sits on the sofa, languid and composed, leaning slightly back; the fabric texture is premium and fine. Beside the sofa is a luxurious black-and-gold marble side table holding an exquisite crystal glass with amber whiskey and a few slowly melting ice cubes, fine condensation beads visible on the glass. Audible: the soft clink of ice against the glass, the faint crackle of a cigar burning, the subtle movement of indoor air, fabric rustling, and restrained breathing.

Reference edit
Images 1–6. Strictly reference example video 1's shot rhythm, transition style, and music.

Keep the buildings, pedestrians, and overall environment in video 1 in realistic-world style, unchanged. Only transform the trees and cars in video 1 into 3D pixel or voxel-block (Minecraft style) builds, with style referenced from image 1. Pixelated objects' motion trajectories must remain normal, and the real environment's shadows and transmitted light must be preserved. Overall effect references video 2.

@image1: the camera first pushes quickly into the coffee surface's milk foam, cocoa powder particles, and dark coffee liquid texture details, letting the surface particles, foam, and liquid ripples gradually fill the screen, occluding everything else. The camera keeps a true macro feel, very shallow depth of field; fine powder drifts slowly in the backlight; the surface texture sits halfway between fine sand and fluid. Avoid frame cracking. At the moment when the coffee powder particles, foam undulations, and liquid vortices closely resemble the dune ridges, wind-erosion texture, and wind-blown sand particles of @image2, transform seamlessly and invisibly into a desert landscape, then continue pushing forward to reveal the full dune vista of @image2. Strictly avoid frame cracking, no black screens, no hard cuts, no obvious VFX, no compositing seams. Overall: real, quiet, restrained — like the same granular material naturally expanding from a coffee micro-surface into a macro desert world. One continuous shot, no stitching in the middle.

Character-action fusion reference
The character's actions, expressions, and performance rhythm in image 1 strictly reference input video 1.
The man stands at the sink on the right side of the frame and hands the washed dishes to the woman on the left, then turns and suddenly flings dish soap foam at the woman at the left edge of the frame with his right hand. Startled, the woman immediately retaliates; the two happily splash foam at each other and dodge, laughing.

Action reference
The character references the actions in video 1 and performs street dance; the character's appearance references image 1 and image 2.

Camera reference
Prompt: referencing the Hitchcock zoom of video 1, make the character in video 2 sing; the singing and performance reference video 3.

Action reference, DIY meme
Reference video 1 for the action imitation. Fixed camera, full shot. Replace the three men in suits with three highly realistic capybaras. The capybaras must strictly follow the original video's motion trajectory: one after another, they quickly lie flat on the floor; then the leftmost capybara jumps to the middle position, the middle one rolls to the far left, the middle one rolls to the far right, the rightmost one jumps to the middle, and finally the middle capybara jumps onto the backs of the other two, stacking into a pyramid. Keep the original fixed camera. Ensure the capybaras' fur texture and lighting blend with the environment.

The character speaks: Follow the wind, live free. Leave worries behind, enjoy the moment. Voice reference: audio 1.

Replace the cat in the video with a dog.

Add a person wearing the same team uniform on the left side of the frame, matching the movements of the others.

Precise character and detail replacement
Replace the child at the back of video 1 with the golden retriever from image 1; change the khaki jacket worn by the leftmost child in video 1 to the denim jacket from image 2.

Remove the green-screen background from video 1 and replace it with a fairy-tale background similar to video 2. Background elements must fully match the character's actions in video 1. Adjust the lighting on the character in video 1 so it completely matches the new background.

Replace the lighting
Reference the video; change the lighting to night.

Live-action scenery replacement
Change the scene outside the window in video 1 to image 1.

Change the line spoken by the woman in video 1 — "We can never be together. It's not that we don't love each other; it's that we truly can't make it to the end." — to the line from audio 1: "Don't go, okay? This time, let's not let go of each other." And slightly adjust the corresponding performance.

Replace the newspaper in the reference video with a book with a green cover; change the chair the person is sitting on to a red sofa; remove the sunglasses the person is wearing, keeping the face clear; remove the car-burning effect so the vehicle returns to a normal state; change the photo the person takes out from their chest to a small black notebook; and add a tree on the left side of the frame.

Reference the video: replace the canned drink appearing at the start with Coca-Cola; change the glowing "FamilyMart" convenience-store sign in the background to a "Hu Hui" sign; replace all the snacks in the plastic bag at the end with Coca-Cola cans; and change the character's final line from "I bought some snacks" to "I bought a bunch of Coke."

Two magicians stand on stage facing the audience, performing a "switch" trick: they both wave their wands at the same time, and a cloud of smoke rises. When the smoke clears, their suit colors have swapped — the man on the left now wears a white suit, and the man on the right now wears a black suit, but their gloves remain the same color. They bow in thanks; the red curtain behind them closes, graduated from deep red to deep blue.

Creative understanding + text effects
In video 1, add the orange-yellow hand-drawn doodle from image 1 around the two people. As the two in video 1 get closer, the doodles multiply, growing from small sparks into a very bright glow. When the two people kiss in video 1, the doodles add pink brushstrokes.