WDCloud · 全能生视频使用说明

先弄清楚怎么选,
再开始生成。

本页只讲五种常用方式

所有示例均为 9:16 竖版。正文保留真实经验、输入素材、完整提示词和生成结果,不用概念包装代替操作说明。

30 秒选模式 ↘
00 / QUICK START

先看手里有什么,再选模式

不需要背功能名称。先确认最想锁定的是想法、资产、开场画面,还是声音。

01 / PARAMETERS

认识参数

先看懂这些设置,可以避免大量无效尝试。

全能生视频参数界面
参数界面示意。下方只解释真正会影响选择和生成时间的项目。
01

清晰度

项目默认为 768p。本页全部示例统一使用 9:16 竖版。

02

质量档位

质量有文戏、动态、超动态三档。普通视频使用文戏即可,能胜任约 90% 的场景;复杂动作和快速打斗再逐档提高。每提高一档,运行时间约增加 10%~20%。

03

细节增强

对细小物体有更强的还原能力,默认不开启。普通模式可以满足大多数场景;开启后,运行时间约增加 10%~20%。

04

总生成时长

例如 8.0s / 60s,表示当前任务已提交 8 秒,所有分段相加不能超过 60 秒。

05

单段长度

单段不超过 20 秒,一般建议控制在 15 秒以内。超过 15 秒时,尾部有很小概率出现画面或声音劣化。

06

历史记录

普通记录保存 3 天、最多 100 条;星标记录长期保存、最多 200 条。取消星标后,从取消时重新计算 3 天。

07

素材库

参考图可以从素材库点击关联到卡槽,也可以直接拖放到对应的参考素材位置。

08

画面比例

支持常见横屏和竖屏比例,画幅本身不影响画质表现。本页案例统一为 9:16,方便直接比较。

段间引导设置界面
多段任务怎么理解

关闭引导是批量提交;开启引导才是连续镜头

段间引导关闭:最多可以提交 8 段,每一段都是独立任务。即使在同一次提交里,也应该作为独立案例查看。

段间引导开启:仅支持 2~3 段,可用于文生视频、全能参考和数字人,让后一段继承前一段的画面状态。

连续长镜头超过 20 秒时,可以考虑开启段间引导。为了提高衔接成功率,要把上一段最后的人物位置、姿态、道具、光线和机位写清楚,并在下一段开头精确继承。

02 / PREPARATION

基础准备

真正决定结果的不是参数数量,而是提示词和资产准备。

提示词助手

不建议直接手写原生提示词

全能生视频对提示词的遵循能力很强,但原生语法比较复杂。具体提示词的写法我们会给予完整教程,可以非常简单地操作,从一个想法就能很快的成片。新手可以先尝试按照此操作里面的案例生成看效果。

经过我们教学后,你只需:

  • 文生视频:直接提交你的想法。
  • 首帧视频:提交首帧图和想法。
  • 首帧音频:提交首帧、驱动音频,并说明音频里说了什么。
  • 全能参考:提交参考图和可选的参考音频;这里的音频只参考音色,不需要说明音频内容。
  • 数字人:提交参考图、驱动音频,并说明音频里的说话内容。
资产

演员、场地、道具和导演

人物资产是演员,场景资产是场地,道具资产是道具,提示词是导演。

不一定要制作过程分镜或完整关键帧。准备好演员、场地和道具,其余部分就可以像导演一样在提示词里调度。

  • 人物资产建议使用四视图。
  • 场景资产建议使用 16:9 横图,明确人物所处环境。
  • 需要强锁定的道具,建议准备多视图。
  • 音频可以作为参考音色,也可以作为驱动音频。
参考音色

只参考“这个声音像谁”

主体说话的音色参考上传音频,但说话内容由提示词决定。

用于:全能参考
驱动音频

直接控制“怎么说、怎么演”

说话内容、节奏、语气、停顿和人物表演由上传音频驱动,最终视频沿用这条音频。

用于:数字人、首帧音频驱动
01
FUNCTION GUIDE

文生视频

一个想法就可以成片

文生视频不需要准备参考图片。无论是演绎片段、口播视频、AB 对话,还是不同风格的片段,都可以先把想法交给提示词助手,再将整理好的提示词用于生成。

这种模式最适合从零开始验证一个想法。人物、场景、对白、音效和镜头组织都写在提示词里;想控制得越细,提示词就要交代得越具体。

9:16 竖版成片 12.3 秒

示例 1 · 从想法直接生成完整片段

宇航员的橙色星球

没有上传参考图,只通过提示词完成场景、角色、对白和镜头揭示。

查看第 1 段完整提示词展开
integrated_multimodal_description:
[Shot 1] Photoreal miniature science-fiction product film on a seamless BLACK studio stage. EXACTLY ONE ten-centimetre astronaut wears a white spacesuit with an orange chest panel and a round clear visor. He stands on a curved orange terrain with highly detailed citrus pores, lit by a hard warm spotlight from upper left and soft cool rim light from the right. Begin at the astronaut's eye height, framing his upper body against pure black. He looks down and shifts his weight once, both boots planted. The astronaut, a young adult male with a clear Mandarin voice and puzzled, restrained delivery (S1), says: <d>[Chinese] 这颗星球,怎么有点不一样?</d>
[Shot 2] At 00:04.000, cut to a low side medium view of the SAME single astronaut. The orange horizon curves sharply away below him; he tilts his helmet and slowly looks toward that curved horizon. Camera makes a small smooth lateral move, maintaining his position at the very top of the sphere. Keep only one person and one continuous orange sphere.
[Shot 3] At 00:08.000, cut wide to reveal the complete arrangement: ONE ordinary orange fruit on ONE smooth white cylindrical display pedestal against a completely seamless black studio cyclorama, with the SAME miniature astronaut standing on the top of the orange. These are the ONLY THREE objects in the entire scene: astronaut, orange, plain cylindrical pedestal. The image is a minimalist luxury product advertisement with spacious black negative space. The astronaut (S1) says with gentle amusement: <d>[Chinese] 原来,我站在一颗橙子上。</d> End with the camera still and the single astronaut looking up. Preserve the clean three-object composition.

overall_soundscape:
Subtle suit movement, soft radio breath and the single specified male voice. Quiet studio background.

non_diegetic_music:
A restrained low science-fiction tone changes to two playful pizzicato notes after the reveal.
9:16 竖版成片 8 秒

示例 2 · 开启细节增强

细微配饰稳定性测试

这个案例开启了细节增强,用来观察戒指、手链和腕表等细小物体在动作中的保留情况。

查看第 1 段完整提示词展开
integrated_multimodal_description:

[Shot 1] Live-action, ultra-realistic cinematic detail test with natural skin texture, high micro-detail preservation, and realistic metal reflections. A beautiful adult Chinese woman around thirty years old stands beside a large window in soft afternoon light. She wears a fitted dark top. Her right hand has one thin silver ring on the index finger, one wider gold ring on the ring finger, a small silver chain bracelet around the wrist, and a compact mechanical wristwatch with a dark leather strap. These four accessories must remain exactly unchanged in number, position, shape, material, and color throughout the entire video.

The camera begins in a tight close-up of her right hand resting on a dark wooden table. She slowly lifts the hand from the table and rotates the wrist toward the camera. At the same time, the camera performs a slow Arc Shot around the hand with small amplitude, keeping the rings, bracelet, watch face, and fingers sharply visible.

At around 00:03.000, she slowly opens and closes her fingers once, then turns the palm slightly downward while the camera continues moving laterally. Reflections move naturally across the metal surfaces, but none of the accessories may disappear, duplicate, change fingers, merge together, or alter shape.

At around 00:05.500, the camera pushes closer toward the watch and rings while she gently brushes one loose strand of hair away from her face using the same hand. The hand then returns slowly toward the table.

The final frame shows the same two rings on the same fingers, the same bracelet, and the same wristwatch fully preserved.

overall_soundscape:
Quiet indoor room tone, subtle clothing movement, light fingertip contact on wood, and natural breathing.

non_diegetic_music:
None.
9:16 竖版总成片 13 秒

示例 3 · 2 段 · 开启段间引导

跨段连续表演

前一段结尾明确写出人物位置、动作、服装和帘子的状态;后一段开头继承同一个可观察状态。

查看第 1 段完整提示词展开
integrated_multimodal_description:

[Shot 1] Live-action realistic late-Ming domestic drama with sensual restraint, intimate character focus, grounded period realism, natural skin texture, realistic silk movement, and predominantly medium-close and close framing. No dialogue, no narration, no distant wide shots.

The main subject is Pan Jinlian, portrayed as a strikingly beautiful adult Chinese woman in her late twenties with refined classical facial features, luminous expressive eyes, realistic smooth skin texture, and a mature alluring presence. Her body is curvy and shapely with a clearly feminine hourglass-like silhouette, full bust, defined waist, and graceful posture. Her long dark hair is arranged in the same elegant late-Ming indoor hairstyle throughout both segments, secured with one delicate hairpin and a few soft strands near the temples.

She wears the exact same fitted light-ivory inner robe and muted-crimson outer robe throughout both segments. The robe has soft silk texture, long flowing sleeves, a defined waist, and restrained embroidery. Her face, hairstyle, body proportions, clothing, makeup, and overall appearance must remain unchanged in the following segment.

The setting is a wealthy late-Ming household corridor leading directly into a private chamber. Red wooden pillars, carved lattice windows, warm late-afternoon light, a thin semi-transparent gauze doorway curtain, and dark wooden floorboards remain spatially consistent across both segments.

A medium-close side tracking shot follows Pan Jinlian walking slowly along the corridor. One hand lightly brushes the wooden railing while her sleeves and robe move naturally with each step. A soft breeze moves the loose strands beside her face.

[Shot 2] At 00:02.000, the shot cuts to a tight three-quarter frontal close-up as she slows near the chamber entrance. She briefly looks toward camera direction with a faint knowing smile, then shifts her gaze toward the doorway.

[Shot 3] At 00:03.500, the shot cuts to a medium-close frontal three-quarter angle. She reaches the doorway and uses her right hand to lift the semi-transparent gauze curtain. The fabric slides across her forearm and rises above shoulder height.

She begins stepping through the doorway. Her left foot crosses the wooden threshold while her right foot remains behind in the corridor. Her torso turns halfway toward the chamber, but her head turns back over her shoulder toward camera direction for one last calm glance.

The final frame must preserve this exact continuity state: Pan Jinlian stands halfway through the chamber doorway, right hand holding the lifted gauze curtain above shoulder height, left foot already across the threshold, right foot still in the corridor, torso turned toward the chamber, head turned back toward camera direction, faint knowing smile still present, loose hair strand beside her cheek, outer robe and gauze curtain still moving gently in the same breeze.

overall_soundscape:
Quiet late-Ming household ambience with faint evening insects, soft footsteps on wooden boards, silk sleeves moving, gauze curtain rustling, distant courtyard birds, and subtle natural breathing.

non_diegetic_music:
None.
9:16 竖版成片 10.1 秒

示例 4 · 单段口播示例

古装人物口播

文生视频也可以直接做口播;需要更长的连续口播时,可以使用段间引导继续生成。

查看第 1 段完整提示词展开
integrated_multimodal_description:

[Shot 1] Live-action, realistic late-Ming-dynasty Chinese domestic drama with sensual restraint, intimate character performance, and grounded period realism inspired by classic old Chinese television historical dramas and traditional chamber staging. One continuous shot only, no cuts. Use a medium-close frontal shot from lower chest upward. The camera remains close to the subject and uses only a very subtle handheld breathing feel, a gentle slow Push In, and an almost imperceptible micro-arc.

The speaker is a beautiful adult Chinese woman around twenty-eight to thirty-two years old with refined classical facial features, luminous expressive eyes, realistic smooth skin texture, delicate makeup, and a mature, confident, teasing presence inspired by a Pan Jinlian-like period-drama archetype. Her body is curvy and shapely with a clearly feminine hourglass-like silhouette, a full bust, defined waist, and graceful posture. Her long dark hair is arranged in an elegant late-Ming updo with a few soft strands framing the face. She wears a fitted muted crimson and warm ivory Ming-style indoor robe with a softly structured neckline, layered silk texture, and restrained embroidery. The clothing should feel sensual through silhouette and movement rather than excessive exposure.

The setting is a wealthy late-Ming domestic chamber at dusk, with dark wooden lattice windows, a carved dressing table, silk curtains, a bronze mirror, warm lantern light, and soft shadow. The atmosphere is intimate, lived-in, and historically grounded rather than fantasy-like.

Her voice S1 is a mature Chinese female voice, soft, slightly husky, intimate, teasing, and controlled, with a subtle smiling undertone and deliberate emotional pauses.

Throughout the whole shot she holds an old traditional Chinese thread-bound book upright at chest height, always visible and mostly facing the camera. The book has an aged kraft-paper-like cover, slightly worn, fibrous, and textured. The front cover displays exactly one large black brush-written Chinese character 男. The book must look like a real old stitched book rather than a modern printed publication. The stitched thread binding appears on one side only along the spine side. The opposite side is the open fore-edge with visible layered paper pages and absolutely no stitching, no thread, and no decorative seam.

She looks directly toward the camera with a knowing amused expression, holding the book lightly against her chest with both hands. She begins in a soft teasing tone S1: <d>[Chinese] 我发现你们都有个习惯。</d>

She pauses briefly, lowers her eyes for a fraction of a second, then looks back up with a faint smile.

She continues S1: <d>[Chinese] 越是在意的东西,越喜欢装作无所谓。</d>

During the phrase about caring too much, she slowly tilts her head by a very small amount and narrows her eyes slightly.

During the phrase about pretending not to care, one fingertip lightly slides along the book cover while the book remains facing camera.

She then raises the book slightly higher and leans toward the camera by only a few centimeters before saying S1: <d>[Chinese] 你们啊,面子比什么都重要。</d>

During the line about face and pride, she lets the final phrase land with a slow amused smile, as if she has already seen through the viewer.

The performance must feel spacious, flirtatious, and natural. Time must be reserved for eye contact, tiny pauses, subtle head movement, and the final teasing hold. Do not rush the delivery.

The final frame preserves this exact continuity state: she remains facing forward, holding the book upright at chest level with the cover visible, chin slightly lowered, shoulders relaxed, and a slow knowing half-smile on her lips.

overall_soundscape:
Quiet late-Ming domestic chamber ambience with faint evening insects outside, soft silk movement, subtle lantern crackle, natural breathing, and minimal hand contact against the book cover. No music. No narration.

non_diegetic_music:
None.
02
FUNCTION GUIDE

全能参考

把演员、场地和道具交给模型

全能参考本质上和文生视频一样,只是增加了图片资产,用来锚定主体和物品的一致性,以及场景的连续性。在这个基础上,演员、场地和道具都可以由你调度。

不一定要先制作完整关键帧。直接准备人物资产、道具资产和场景资产,通常更容易生成,也更灵活。想要怎样的表演方式、人物关系和镜头安排,在提示词里明确即可。

场景参考不是必须上传;如果对环境没有严格要求,可以直接在提示词中描述。有场景参考时,方向性和确定性会更强。

9:16 竖版成片 6.6 秒

示例 1 · 人物资产 + 道具资产

人物与道具的调度

只提供演员和道具,场景、光线和动作由提示词安排。

提交的图片资产
人物资产
人物资产下载原图
道具资产
道具资产下载原图
查看第 1 段完整提示词展开
subject_definitions:
<Subject 1> is the adult Chinese woman from the four-view identity sheet, center-parted black hair in a low bun with loose face-framing strands, white cross-collar wide-sleeved robe and white cloth belt, loose white trousers and off-white shoes, referenced from <Picture 1>.
<Subject 2> is the small sculptural lamp: an amber glass sphere suspended inside two intersecting brushed brass rings on a short dark round pedestal, with no writing or controls, referenced from <Picture 2>.

summary:
[reference generation] An encounter built from reusable visual references.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - preserve facial identity, hairstyle, body proportions and clothing; do not copy the multi-view layout or neutral studio background.
<Subject 2> (appears in [Shot 1]): fully_preserved - preserve the sphere, two-ring geometry, brass finish and pedestal proportions; the reference studio background is not the setting.

detailed_description:
Photorealistic cinema, real skin and textured fabric, selective depth of field, a single full-frame vertical composition, no captions or book props. 
[Shot 1] <Subject 1> stands behind a waist-high basalt plinth in a dark circular hall; <Subject 2> sits centered on the plinth, its entire pedestal visible. Start waist-up with her on the left and the lamp on the right. The amber sphere brightens gradually, casting moving warm reflections on her white sleeve; she leans forward slightly and raises her eyes with surprise. Camera arcs right a quarter circle, keeping the two-ring silhouette readable and her face in focus. Cool overhead skylight separates her hair from the dark hall. No contact with the rings, no extra limbs or duplicate people.

overall_soundscape:
Soft footsteps, wind and restrained resonant mechanical tones from the lamp when visible; speech only as explicitly staged above.

non_diegetic_music:
N/A
9:16 竖版成片 12.3 秒

示例 2 · 多人物参考 + 道具参考

两位演员与一件道具

图片资产用于锁定两位人物和道具;人物说什么、场景在哪里以及镜头怎样切换,仍由提示词控制。

提交的图片资产
人物资产 1
人物资产 1下载原图
人物资产 2
人物资产 2下载原图
道具资产
道具资产下载原图
查看第 1 段完整提示词展开
subject_definitions:
<Subject 1> is the elderly Chinese man from the four-view identity sheet: deeply wrinkled face, long white beard, white eyebrows, white hair in a small topknot without a projecting hairpin, black cross-collar robe, black sash, loose black trousers and black cloth shoes, referenced from <Picture 1>.
<Subject 2> is the adult Chinese woman from the four-view identity sheet, center-parted black hair in a low bun with loose face-framing strands, white cross-collar wide-sleeved robe and white cloth belt, loose white trousers and off-white shoes, referenced from <Picture 2>.
<Subject 3> is the small sculptural lamp: an amber glass sphere suspended inside two intersecting brushed brass rings on a short dark round pedestal, with no writing or controls, referenced from <Picture 3>.

summary:
[reference generation] An encounter built from reusable visual references.

retention_analysis:
<Subject 1> (appears in [Shot 1], [Shot 2], [Shot 3]): fully_preserved - preserve facial identity, hairstyle, body proportions and clothing; do not copy the multi-view layout or neutral studio background.
<Subject 2> (appears in [Shot 1], [Shot 2], [Shot 3]): fully_preserved - preserve facial identity, hairstyle, body proportions and clothing; do not copy the multi-view layout or neutral studio background.
<Subject 3> (appears in [Shot 1], [Shot 2], [Shot 3]): fully_preserved - preserve the sphere, two-ring geometry, brass finish and pedestal proportions; the reference studio background is not the setting.

detailed_description:
Photorealistic cinema, real skin and textured fabric, selective depth of field, a single full-frame vertical composition, no captions or book props. 
[Shot 1] In a weathered open pavilion at blue hour, <Subject 1> stands screen-left, <Subject 2> screen-right, with <Subject 3> stationary on a stone table between them. A warm practical lamp off camera-left and the amber sphere light their faces, cool dusk behind. Medium two-shot establishes the left-right relation. <Subject 1>, an elderly man with a low slightly rough Mandarin voice and measured pace (S1), says: <d>[Chinese] 你看见的,是光,还是答案?</d> Only his mouth moves; the woman listens with closed lips.
[Shot 2] At 00:05.000, the camera cuts to an over-shoulder medium close view of <Subject 2> from the same side of the dialogue axis, the lamp still at the lower edge. The woman with a clear adult female Mandarin voice and calm deliberate delivery (S2) replies: <d>[Chinese] 我看见的,是另一种可能。</d> The man keeps his lips closed.
[Shot 3] At 00:09.500, the camera cuts back to the two-shot, both silent, the amber sphere illuminating one step brighter as they exchange a tiny smile. No ring manipulation or reference-sheet layout.

overall_soundscape:
Soft footsteps, wind and restrained resonant mechanical tones from the lamp when visible; speech only as explicitly staged above.

non_diegetic_music:
N/A
03
FUNCTION GUIDE

数字人

不只是固定机位口播

数字人其实是全能参考的变种。人物资产、道具资产和场景资产都可以使用,最主要的区别是声音由音频硬驱动,人物表演的节奏和情绪也跟随音频。音频是什么,最终视频使用的就是什么。

这种方式的好处是,可以提前准备质量较高、表现力更强的语音;在多个生成任务中,也能保持声音特点、情绪和表达节奏的一致性。音频可以由 IndexTTS 生成。

数字人仍然是一种很灵活的表演方式。它不只适合正对镜头说话,也可以让人物处在不同环境中行动和表演,只是需要提前多准备一步驱动语音。

9:16 竖版成片 9.5 秒

示例 1 · 人物与场景参考 + 驱动音频

山亭里的黑袍老人

驱动音频直接决定人物说话内容、停顿、节奏和情绪,最终视频沿用这条音频。

提交的图片资产
人物资产
人物资产下载原图
场景资产
场景资产下载原图
提交的驱动音频
驱动音频下载 WAV
查看第 1 段完整提示词展开
subject_definitions:
<Subject 1> is the elderly Chinese man shown from four views in <Picture 1>: white hair in a small topknot without a projecting hairpin, long white beard, white eyebrows, deeply wrinkled face, black cross-collar robe, black waist sash, loose black trousers and black cloth shoes. The four views depict one person only.
<Subject 2> is the empty mountain pavilion environment from <Picture 2>, with weathered dark timber columns, damp gray stone paving, a wall opening, stone railing and a misty valley lit from the right.
<Audio 1> is the authoritative synchronized speech track for <Subject 1> (S1), reused as the final audio track.
summary:[reference generation + audio reuse] <Subject 1> speaks to the viewer within <Subject 2>, with performance and mouth movement driven by <Audio 1>.
retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain facial identity, age, topknot without hairpin, beard, body proportions and black clothing; the reference-sheet layout and gray studio background are not part of the scene.
<Subject 2> (appears in [Shot 1]): partially_preserved - retain the pavilion materials, wall opening, open valley side and morning light direction; adapt camera placement to a vertical portrait composition, without copying the wide reference framing.
<Audio 1>: fully_copy - <Audio 1> is reused 1:1 as the target video's complete final audio track.
detailed_description:
Photorealistic Chinese period cinema, tactile weathered materials, charcoal and slate tones with restrained warm side light, natural skin texture, no beauty filter. A single vertical image fills the frame; no books, pages, writing, captions, collage or split screen.
[Shot 1] <Subject 1> (S1) speaks in exact synchronization with <Audio 1>, with lips, jaw, facial muscles, breathing, pauses and emphasis following the source audio exactly:<d>[Chinese] 没有改变命运的野心,没有坚如磐石的决心,这本旷世奇书,你千万不要碰。</d>Begin in a chest-up view with the man on the left half of the vertical frame, the valley soft behind his right shoulder. During (没有改变命运的野心), he holds the viewer with a steady gaze. During (没有坚如磐石的决心), he makes one slight downward nod. At the pause before (这本旷世奇书), he becomes still. During (你千万不要碰), his gaze stays fixed and the jaw settles naturally at the end. The camera makes a small lateral drift in the first half, then stops for the final warning. Preserve the exact pauses in the audio; do not add a pause or speed up delivery. The man stands in <Subject 2>; low sunrise from camera-right adds a narrow warm edge to his cheek and shoulder, with soft cool fill from the valley. Focus stays on his eyes and mouth, the background softened. Gentle wind moves only the ends of his beard and loose fabric. Natural blinking follows the speaking performance. No large body sway, mouth occlusion or extra speaking person.
noverall_soundscape:
The complete final soundtrack is the unaltered <Audio 1>; do not add new speech or environmental sound over it.
non_diegetic_music:\nN/A
9:16 竖版成片 10.8 秒

示例 2 · 更换驱动音频与表演场景

骑鹤飞行中的讲述

同一个人物资产可以放进完全不同的场景。数字人并不只适合固定机位口播,也可以完成更灵活的表演。

提交的图片资产
人物资产
人物资产下载原图
提交的驱动音频
驱动音频下载 WAV
查看第 1 段完整提示词展开
subject_definitions:
<Subject 1> is the on-screen speaker shown in <Picture 1>; preserve his elderly Chinese male identity, white hair tied in a topknot, long white beard, weathered facial features, dark traditional robe, body proportions, and distinctive visual details from <Picture 1>.
<Audio 1> is the complete source spoken-audio track performed by <Subject 1> (S1) and is reused as the target video's synchronized final spoken-audio track.

summary:
[reference generation + audio reuse] The target video shows <Subject 1> performing a continuous close-range speaking sequence synchronized to <Audio 1> while seated on a flying crane, with continuously changing scenery rushing past in the background.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain his facial identity, age appearance, hairstyle, beard shape, clothing, body proportions, and distinctive visual details established by <Picture 1>.
<Audio 1>: fully_copy - <Audio 1> is reused 1:1 as the target video's complete final spoken-audio track.

detailed_description:
The target video follows the visual language of classic late-1980s to 1990s Hong Kong fantasy-wuxia live-action cinema, especially the soaring romantic fantasy energy of Tsui Hark and Ching Siu-tung, with rich atmospheric color, lyrical motion, and dramatic scenic transformation. The shot should feel like a Hong Kong fantasy film rather than generic “cinematic” footage. Use practical-looking live-action texture, drifting mist, strong wind, layered mountains, changing sky color, and vivid moving backgrounds. The palette should be expressive and saturated but still naturalistic: sunset gold, emerald green, deep cyan distance haze, weathered stone gray, dark robe black, and warm amber light on the face. Preserve a real live-action look throughout; do not drift into game CG or glossy synthetic fantasy render.

[Shot 1] Use one uninterrupted shot. Frame <Subject 1> in a close near-full-body or large half-body composition, approximately from mid-thigh or waist upward, so his face remains large and readable while enough of the crane's back, neck base, and moving flight posture stay visible to sell the action. Keep the mouth region fully visible and unobstructed throughout. Use a slightly low three-quarter front angle, as if the camera is flying alongside and just ahead of him in the air. Keep <Subject 1> slightly off-center rather than dead center, with the crane body anchoring the lower frame and the changing world sweeping behind him.

<Subject 1> sits astride a large white crane in flight high above mountains. The crane is elegant, powerful, and continuously gliding through the sky with steady wing rhythm and believable body motion. <Subject 1> remains seated firmly and naturally on the crane's back, one hand lightly stabilizing his posture on the crane or resting near his thigh, while the other hand is free for small speaking gestures. His robe, sleeves, beard, and hair move continuously in the wind. The crane's motion should feel fluid and coherent, with subtle vertical buoyancy and forward speed, not mechanical hovering.

The background must remain highly dynamic throughout the full speech. As the crane flies, the scenery continuously changes behind and around him: layered mountain ridges, drifting clouds, misty cliffs, temple roofs on peaks, waterfalls, ancient stone pavilions, forests, river valleys, and changing sky depth all sweep past in a flowing progression. The viewer should feel constant forward travel. Some sections may pass through cloud haze, some above forested ridges, some near temple silhouettes, and some against open glowing sky. The moving environment should feel rich and cinematic, with strong parallax and continuous visual change, but <Subject 1>'s face must remain readable and stable.

The camera must not perform a long push-in. Keep the shot alive with a restrained floating motion: slight aerial handheld breathing, subtle responsive drift, and a gentle small-amplitude side-following arc as if a second flying camera is tracking alongside the crane. The motion should feel elegant, alive, and wind-borne, not flashy or chaotic. The expressive power comes from the moving space, the Hong Kong fantasy atmosphere, the crane flight, and the changing scenery rather than from a simplistic push toward the face.

<Subject 1> (S1) performs the spoken content carried by <Audio 1> and says in synchronization with <Audio 1>: <d>[Chinese] 然而历经千年,无数智者穷经皓首,试图拨开迷雾,直到唐朝,一部旷世经典才应运而生。</d>

His visible speaking performance remains tightly synchronized to <Audio 1>, with natural mouth articulation, jaw movement, breathing rhythm, pauses, and subtle expression changes following the exact cadence of the source performance. Even while flying, his delivery should remain calm, weighty, and quietly revelatory, like an old sage speaking above the world.

Facial performance should remain restrained and wise. Maintain calm eye focus generally toward the camera or slightly through it, with natural blinking, subtle eyebrow changes, slight chin dips, and small controlled head movement. If gestures are used, keep them measured and readable: one or two restrained explanatory gestures near the torso or outward from the sleeve, never exaggerated. He should feel grounded and unhurried despite the dramatic flight.

A key requirement is dynamic continuity. The crane flight, robe movement, beard movement, hand gestures, speech, and constantly changing background should all feel smoothly connected from beginning to end. Avoid jitter, broken continuity, deformed limbs, unstable hands, frozen crane motion, or repetitive looping scenery. Keep <Subject 1>'s face clearly identifiable and in optical focus throughout. Preserve hairstyle, beard shape, robe silhouette, facial structure, and body proportions consistently from beginning to end.

The shot should end with the crane still in active forward flight, the scenery still evolving, and <Subject 1> settling naturally after the final phrase in a composed seated posture, leaving the impression that the journey is still continuing beyond the end of the segment.

overall_soundscape:
<Audio 1> is reused as the exact final spoken-audio track, preserving its original timing, phonemes, pauses, breaths, and performance identity. Underneath it, keep only a very subtle environmental layer of high-altitude wind and distant open-air atmosphere, without interfering with speech clarity.

non_diegetic_music:
N/A
04
FUNCTION GUIDE

首帧视频

先把开场画面锁定

首帧视频和文生视频本质相同,只是片段的第一帧已经被锁定,之后的所有画面围绕这张首帧继续衍生。

它适合已经确定人物、场景、构图和起始姿态的情况。首帧可以来自 Image2,也可以是已有素材;后续动作、镜头和声音仍然由提示词控制。

9:16 竖版成片 6.6 秒

示例 1 · 锁定首帧后继续演绎

月门被光唤醒

片段从给定画面开始,之后的动作、光线和镜头围绕首帧继续发展。

提交的图片资产
首帧图
首帧图下载原图
查看第 1 段完整提示词展开
For the target video, at 0.00 seconds into the target video, Picture 1 (from [Shot 1]) is fully referenced.

integrated_multimodal_description:
[Shot 1] Photorealistic cinema, real skin and textured fabric, selective depth of field, a single full-frame vertical composition, no captions or book props. Begin exactly from the reference woman and stone moon gate. She slowly shifts her eyes toward the circular opening, then turns her shoulders toward it; her hands stay low. Warm amber motes ignite along the existing inner circumference. Camera pulls back two metres to expose the terrace over the clouds. Keep the gateway solid, the woman single and the sunlight from the same side. Her robe hem responds to a brief gust; end on her stillness facing the illuminated opening.

overall_soundscape:
Grounded location sound, cloth movement and wind matching the visible action; no intelligible speech.

non_diegetic_music:
N/A
05
FUNCTION GUIDE

首帧音频驱动

一张图片加一条语音就能开始表演

这个模式是首帧视频的变种,最主要的区别是人物语音由驱动音频控制。它可以同时锁定场景、人物和语音,特别适合用一张图片快速生成数字人表演。

不要只把它理解为固定镜头口播。也可以把人物当成固定场景里的演员:先把姿态、构图和机位调好,再由音频驱动台词、节奏和表演。

9:16 竖版成片 10.7 秒

示例 1 · 首帧锁画面 · 音频锁表演

雪院里的讲述者

只提交一张首帧图和一条驱动音频,人物就可以开始表演;也可以把它理解为固定场景里由音频驱动的演员。

提交的图片资产
首帧图
首帧图下载原图
提交的驱动音频
驱动音频下载 WAV
查看第 1 段完整提示词展开
A continuous live-action medium close-up begins exactly from <Picture 1>. Preserve the elderly Chinese man's identity, facial structure, white hair, long white beard, black traditional robe, snowy courtyard, architecture, natural skin texture, composition, lighting and color appearance from the first frame.

The elderly man faces the camera and speaks visibly in exact synchronization with <Audio 1>. His mouth, lips, jaw and facial muscles naturally follow the syllable rhythm, pauses, breaths, emphasis and emotional progression of the original Mandarin speech.

He begins with a solemn and contemplative expression. While speaking, he maintains direct eye contact with the viewer. His already-raised right hand continues one restrained, natural explanatory gesture. On “直到唐朝”, his eyes become slightly more focused and the right hand gently emphasizes the turning point. On “一部旷世经典才应运而生”, his expression shifts subtly from searching and uncertainty to calm conviction, then his hand settles naturally.

The falling snow continues gently in the background and on his robe. Natural blinking, breathing and subtle head movement. Keep the face and mouth clearly visible throughout. Locked and stable camera, continuous single shot, no cuts, no sudden camera movement, no profile turn, no exaggerated gestures, no hands covering the mouth, no additional people, no subtitles and no visible text.

The elderly man (S1) says exactly once:
<d>[Chinese] 然而历经千年,无数智者穷经皓首,试图拨开迷雾,直到唐朝,一部旷世经典才应运而生。</d>

<Audio 1> is the authoritative speech source. Do not add, omit, repeat, rewrite, extend or regenerate any spoken words. After the final syllable of “生”, his mouth closes naturally and he maintains a calm, steady expression.
DEEP PRACTICE

深入练习区

这里保留此前制作的其他真实案例。与正文重叠的主案例已经移除;原本在一次任务中批量提交、但没有开启段间引导的多段结果,全部拆成独立案例展示。

A01 · 独立案例

雨夜车站的透明鲸

倒影揭示尺度变化声画设计

A04 · 独立案例

山亭背后的悬空城

首帧人物前后景揭示电影式切镜

A13 · 独立案例

多参考数字人口播

四视图锁人音频驱动场景由文字生成

A14 · 独立案例

海底讲述者

虚构场景跨切镜口播环境与人物分层

原任务 B01 · 独立案例 1

一次任务两个世界 · 第 1 个独立结果

一次两段独立题材无需场景连续

原任务 B01 · 独立案例 2

一次任务两个世界 · 第 2 个独立结果

一次两段独立题材无需场景连续

原任务 B02 · 独立案例 2

同一讲述者跨山海 · 第 2 个独立结果

两段独立音频共用人物资产场景切换

原任务 B03 · 独立案例 1

人和光球跨世界 · 第 1 个独立结果

跨段复用人物物品双锁定独立场景

原任务 B03 · 独立案例 2

人和光球跨世界 · 第 2 个独立结果

跨段复用人物物品双锁定独立场景

C01 · 开启段间引导

没有参考图也能续场

纯文生连续明确衔接状态持续运镜

C02 · 开启段间引导

资产加引导的连续奇观

资产一致性前段尾部引导场景状态延续

C03 · 开启段间引导

口播跨段不断场

分段语音场景接续持续口播表演