✦

wan/3-0-video

Name: Wan 3.0 Video API
Brand: Wan

可商用Wan 系列模型之一 —— 浏览全部模型 →

⚡ 通过 API 调用

Wan3.0 是阿里巴巴新一代全能视频生成模型，专为多模态创作打造。它支持文本、图片、视频、音频和结构化参考，生成更一致、更真实、更可控的视频内容，叙事和编辑能力更强。

定价： 480p 起 10.4 积分 / 秒（5 秒 52 积分），按输出秒数计。各分辨率见下方费率表。

在线试跑

表单已按模型规格预填 —— 可直接运行，也可先改任意字段。

可用

首尾帧与参考素材互斥；切换会清空另一组。文档与网页链接也不能同时填写。

frame_key

上传图片

提示词

Generate an approximately 30-second, 16:9 horizontal high-aesthetic urban MG chase short film. Image1 @Image1 is used as the reference for the city space, neon atmosphere, road structure, and overall visual texture; Image2 @Image2 is the only reference for the male character appearance; Image3 @Image3 is the only reference for the female character appearance. Strictly maintain the consistency of both characters' faces, hairstyles, body shapes, clothing structures, color palettes, and visual identity.
【Overall Style】
High-end motion graphics dynamic illustration style, combining flat MG graphic language, refined 2.5D spatial layering, and restrained 3D volume. The overall style should not be a realistic live-action film, not childish cartoon style, and not messy heavy industrial cyberpunk. The atmosphere should be stylish, cool, beautiful, tense, and eye-catching, like an action-movie-level urban chase promotional short film. The camera language should heavily use handheld tracking shots, rapid whip pans, snap zooms, low-angle ground-level shots, backward tracking shots, whip transitions, and occasional drone wide shots, creating a strong sense of pursuit pressure.
【Character Action Logic】
The female character's advantage is lightness, speed, agility, and using the environment and route changes to escape. The male character's advantage is strength, prediction, pressure, and interception ability. Both characters are martial arts experts, so the confrontation should not be chaotic grabbing, but a battle around the glowing ball involving "grabbing, blocking, redirecting force, switching hands, using momentum, turning, dodging, and counterattacking." The female character often uses turning movements, force redirection, route changes, and environmental obstacles to create distance; the male character often uses early positioning, wrist interception, body blocking, route sealing, and close-range suppression.
【Glowing Ball】
The glowing ball can be held in one hand, with a white core and soft purple-pink-blue halo. It is the only target being fought over throughout the entire film. The glowing ball should realistically illuminate the hands, facial edges, clothing, and surrounding environment. There is always only one glowing ball. Its size remains consistent. It never duplicates, disappears, or deforms. During movement, it only creates short, clean light trails.
【0.0-4.0 seconds | City traversal and female character introduction】
The camera begins with a high-speed flight through a city nightscape, passing through skyscrapers, neon signs, and road lights, then diving along the main road into the urban district, crossing traffic, and rapidly lowering to street level. As the camera suddenly stops, the female character rushes into frame from the side, tightly holding the glowing ball.
First show a series of fast close-up shots: footsteps, the hand gripping the ball, her backward glance, and her forward-running upper-body posture. She is not casually running; she is clearly escaping, quickly observing both the route ahead and behind her, showing that she knows someone is chasing her.
【4.0-8.0 seconds | Male interception, first close-range confrontation】
When the female character enters a narrow street passage, the male character cuts in from another route ahead, showing that he is not blindly chasing but predicting and intercepting.
After seeing the male character, the female character does not stop immediately. Instead, she attempts to diagonally change direction and bypass him. The male character immediately reaches out and grabs the wrist holding the glowing ball, with the clear goal of taking the ball.
The female character instantly pulls her arm back to protect the ball, while using the pulling force to rotate her body around the outside of his grabbing hand, creating the first force-redirection movement. The male character immediately switches his other hand toward the ball route and uses his shoulder to block her escape path.
The female character lowers her center of gravity, slides underneath his arm, and before escaping position, lightly blocks his chest with her forearm to create half a step of space. The male character catches up again and briefly touches the ball but fails to secure it.
The camera stays close to both characters, using handheld close-ups, rapid zooms, and small circular movements, always focusing around the glowing ball and hand movements. Finally, the female character performs a fast turn to escape the male character and continues running forward.
【8.0-13.0 seconds | Entering the crowd, female character uses the environment】
The female character rushes into a crowded area. She does not run randomly, but intentionally chooses a complex environment because she is more agile than the male character.
She first moves sideways between two pedestrians, then uses roadside stalls and railings to quickly change direction, making it difficult for the male character to approach directly.
The male character continues chasing through the crowd without slowing down, occasionally pushing through shoulders to create pressure. The camera mainly uses wide-angle handheld close-follow shots. The frame can be briefly blocked by pedestrians, creating a realistic chaotic feeling.
At one point, the male character grabs the back of the female character's coat. The female character does not pull backward forcefully. Instead, she uses the pulling force to rotate sideways, redirects the force, then uses her elbow to push away his forearm before slipping past his side.
The two have their second close confrontation inside the crowd: the male character grabs the female character's forearm with one hand and reaches for the glowing ball with the other. The female character raises the ball-holding hand to avoid the grab, locks his wrist with her other hand, rotates her body in the opposite direction, and uses momentum to pull him slightly off balance.
The female character slips away from his side and escapes again.
【13.0-16.5 seconds | Entering traffic, switching to bicycles】
After escaping the crowd, the female character reaches the edge of the main road. She realizes continuing on foot will allow the male character to catch up, so she immediately changes transportation.
She grabs the handlebar of a bicycle parked nearby and uses her forward momentum to jump onto it, immediately beginning to ride without stopping.
The male character exits moments later, quickly grabs another bicycle and continues the pursuit.
Use a brief wide shot to establish the space: city skyscrapers, main roads, traffic flow, and both characters' routes. Then immediately return to a low-angle perspective and enter a high-speed bicycle chase.
【16.5-24.0 seconds | Bicycle chase through traffic and third confrontation】
The female character's cycling route is more flexible. She rides close to cars, changes lanes, leans the bicycle, and cuts diagonally through traffic gaps to create distance.
The male character's route is more direct and aggressive, constantly reducing the straight-line distance.
The camera alternates between ground-level tracking shots, backward tracking shots, high-speed side movements, and brief aerial perspectives.
The traffic should react realistically to the chase: when the characters pass near car fronts, vehicles brake suddenly, lights turn on, horns sound, and drivers slightly avoid them, creating a sense of danger.
During the bicycle chase, the male character first reaches side-by-side and reaches for the glowing ball. The female character does not immediately pull away. Instead, she intentionally keeps the ball slightly outward for half a beat, causing the male character to extend his body further.
When his balance shifts, the female character suddenly leans the bicycle and changes direction, causing him to miss.
During the second approach, the male character becomes smarter. He uses a slowing vehicle to block the female character's escape route, then approaches from the side rear and almost grabs her wrist holding the ball.
This time the female character does not resist directly. She quickly switches the ball to the other hand, uses the bicycle handlebar for a half-turn movement, and slides away from the male character's reach.
The two bicycles nearly collide. The camera should create strong speed and danger.
【24.0-27.5 seconds | Final pressure and female escape】
The traffic becomes more complicated at the intersection ahead.
The male character uses his straight-line speed advantage again and nearly reaches the female character's shoulder from behind.
The female character notices a large bus crossing horizontally and a line of suddenly stopping vehicles, instantly recognizing an escape opportunity.
She intentionally maintains her original route to keep the male character locked onto her position. When he gets close, she suddenly leans the bicycle and cuts through the narrow gap between the front of a braking car and the bus.
The male character reacts slightly slower and is forced to slow down and detour because of the angle between the car and bus.
The camera quickly pans and becomes completely blocked by the side of the bus, creating a natural transition.
【27.5-30.0 seconds | Ending】
After the bus passes, the male character rushes out but discovers that the female character has already created distance.
In the distant neon street corner, only a rapidly moving glowing ball remains visible.
The camera quickly pushes toward the distance. The female character briefly looks back in the deeper neon street, showing a slightly proud and relaxed expression, then continues riding away with the glowing ball.
Finally, the camera slowly rises and pulls away, revealing the busy night city, continuous traffic flow, and the disappearing light point in the distance. The female character successfully escapes.
【Camera Movement Requirements】
70% of the entire film should have handheld camera feeling. Do not make it overly stabilized.
Close-range confrontation scenes should mainly use handheld shots, rapid push-ins and pull-outs, and small circular movements.
Crowd chase scenes should strengthen obstruction, blocking, and breathing rhythm.
Traffic chase scenes should emphasize low-angle shots, opposing movement, rapid zooms, and sudden braking atmosphere.
Wide shots are only used to establish space and provide pacing breaks.
The camera should always serve the ownership of the glowing ball, the distance between the two characters, and the direction of pursuit. Do not add meaningless flashy rotations.

Text prompt used to describe the desired video content. Supports both Chinese and English, with a maximum length of 20,000 characters.

参考图

添加图片

该变体使用 3 张输入素材。

Maximum 10 images. Mapped in array order to Image1, Image2, ... in the prompt. Specifications are the same as first_frame_url. Cannot be used together with first frame / last frame.

参考视频

添加视频

Maximum 5 clips. Each clip: 1–15s; total combined duration ≤ 15s. Mapped in array order to Video1, Video2, ... in the prompt. Formats: mp4, mov. Single side: [240, 4096] px; aspect ratio ≤ 8:1; file size per clip ≤ 100MB. Additionally, on the output side: total input video duration + output duration (duration parameter) must not exceed 30 seconds.

参考音频

添加音频

Maximum 5 clips. Each clip: 1–15s; total combined duration ≤ 15s. Mapped in array order to Audio1, Audio2, ... in the prompt. Formats: wav, mp3; file size ≤ 15MB per clip. When not used as the sole media input, it is still recommended to pair with images or videos.

reference_link_urls · 该字段是嵌套结构，暂未渲染（PlaygroundArrayForm）

Maximum 1 publicly accessible webpage (no login required). Cannot be used together with reference_file_urls, nor with first frame / last frame.

reference_file_urls · 该字段是嵌套结构，暂未渲染（PlaygroundArrayForm）

Maximum 1 file. Cannot be used together with reference_link_urls, nor with first frame / last frame. Supported formats: docx, doc, xlsx, xls, pptx, ppt, pdf, txt, key, pages, numbers, md. File size ≤ 100MB. For file types such as pdf, docx, ppt, key, pages, etc., page count ≤ 50 pages.

分辨率

Preset resolution levels for video generation.

画面比例

Aspect ratio for generated video. "Adaptive" auto-recommends a ratio based on input media proportions and intent.

时长5

Duration of the generated video, in seconds. When no video input is provided, the value must be an integer in the range of [2, 30]. When a video input is provided, the total duration of the input video plus the output video must not exceed 30 seconds. Passing -1 enables smart duration mode, where the model automatically recommends an appropriate duration based on the input.

音频

Whether the output video includes audio. true (default): Includes audio; false: No audio track.

seed0

Random seed for result reproducibility. 0 <= x <= 2147483647

nsfw_checker

A configurable parameter. Defaults to true in the Playground.

余额： —本次消耗208 积分

结果

历史

该变体的示例输出 —— 点「运行」生成你自己的。

Wan 的更多模型

Wan3.0 API 重新定义多模态视频创作，控制力更强

体验视频生成的下一步进化：Wan3.0 API 结合更长的叙事、更强的参考一致性、精确编辑和沉浸式视听创作，让视觉叙事更完整。

用 Wan3.0 API 从更丰富的多模态输入开始视频创作

Wan3.0 是阿里巴巴最新的全能视频生成模型，面向更长的叙事、更丰富的参考理解、更强的视觉一致性和更精确的创作控制。Wan3.0 API 可以基于不同类型的输入素材进行创作，包括文本、图片、视频、音频、关键帧、文档和网页，通过更灵活的参考来塑造场景、动作、声音、视觉特征和叙事方向。最新支持的输入、生成模式和 API 用法，请参阅 ModelFold 上的 Wan3.0-Video API 文档。

把文字提示词变成完整的视频构思

文字能做的远不止描述一个简单场景。借助 Wan 3.0 API，书面指令可以塑造角色、环境、运镜、动作逻辑、光影、节奏、氛围，甚至视频的情绪起伏，让详细的提示词成为结构化创作的灵活起点。如果多模态视频流程已经够用，就继续用 Wan 3.0 Video；如果需要视频生成同时带原生立体声，可以换用更新的 MiniMax H3（Hailuo-03）API。

用图片参考把视觉特征带进动态画面

图片参考能在动作开始前就定下视频的样子。角色外观、服装、道具、地点、材质、构图和整体风格都可以通过图片来引导，让 Wan3.0-Video API 把可辨认的细节带进动态场景，连贯性更强。

用视频参考引导动作和场景流动

有些创作意图演示比解释更容易。现有素材可以提供动作语言、表演提示、节奏、镜头行为和场景动态，帮助 Wan3.0 API 以更可控的动作和更清晰的流动感生成或改造视频。

让声音方向与画面生成联动

音频对场景感受的影响不亚于画面。语音、音乐、节奏、人声和其它声音参考，为 Wan 3.0 Video API 提供关于节奏、情绪、表演能量和音画配合的更多上下文，让最终效果更统一。

用关键帧让场景推进保持一致

关键帧为片段的开始、发展和结束提供强有力的视觉锚点。通过首帧或尾帧参考，Wan3.0 API 能引导主体连贯、转场逻辑和场景发展，让生成的视频方向更一致。

把文档和网页带进创作流程

创作素材不一定从提示词开始。简报、演示文稿、表格、PDF、文本文档和网页也能提供有价值的结构和上下文，让 Wan 3.0 API 把现有信息变成视频生成流程的一部分，而不是单独的准备工作。

了解 Wan3.0 API 为何是视频创作的一次大飞跃

Wan3.0 API 全能多模态创作

Wan3.0 把更丰富的创作上下文融入生成过程，让视频创作不再局限于常规提示词。图片、视频、音频、文档、网页等不同类型的参考素材，都能帮助引导场景发展、视觉一致性和创作方向。最新支持的输入类型、可用功能和 API 用法，请参阅 ModelFold 上的 Wan3.0-Video API 文档。

Wan 3.0 API 原生 30 秒叙事

更长的原生生成让场景有更多空间交代背景、展开动作、在故事节点间转场，并收出更完整的结局。Wan 3.0 API 支持最长 30 秒的视频，并能根据内容智能调整时长，为叙事结构和创作自由留出更多空间。

Wan3.0-Video API 像素级参考一致性

在动作和场景变化中，角色、服装、道具、环境、音频线索和空间关系都能与原始参考保持高度一致。Wan3.0-Video API 加强了细节还原和连贯性，需要保持视觉特征稳定时，基于参考的创作更可靠。

Wan 3.0 Video API 精确视频编辑

通过详细的创作指令改造现有视频，而不必改动原素材的全部内容。Wan 3.0 Video API 支持对主体、背景、环境、光影、视觉风格等元素做精确调整，还支持参考引导编辑和视频延长，创作打磨更灵活。

Wan3.0 API 沉浸式视听生成

视觉叙事搭配更有表现力的音频生成，让场景更完整。Wan3.0 API 提升了自然音色、地方口音表达、音乐节奏、环境音和音画同步，让动作、表演、氛围和声音融为一体。

Wan3.0-Video API 电影级画面与动作控制

Wan3.0 在光影、色彩、构图、角色表演、镜头语言和场景动态上提升了真实感和美学控制。Wan3.0-Video API 能以更高的视觉精度处理复杂的动作编排、真实的物理交互、细腻的材质和微妙的表演，效果更电影化。

如何在 ModelFold 上用 Wan3.0-Video API 开始创作视频

1
设置 ModelFold 账号，开通 Wan3.0-Video API
注册 ModelFold 账号或登录已有账号，即可开始使用 Wan3.0-Video API。拿到 API 凭据后，配置请求环境，准备好把视频生成能力接入你的应用所需的设置。
2
为 Wan3.0-Video API 请求准备创作素材
把提示词与支持的创作参考和指令组合起来，创建视频生成请求。Wan3.0-Video API 能处理不同形式的创作上下文，帮助引导视觉一致性、场景发展、动作设计和整体叙事方向。详细的输入选项和请求格式请参阅最新的 ModelFold 文档。
✓
提交 Wan3.0-Video API 任务并获取结果
通过 ModelFold 视频接口向 Wan3.0-Video API 发送生成请求。任务完成后响应里直接返回视频地址，无需轮询，即可把结果接入你的创作项目或应用。

通过 Wan 与 Seedance 探索 AI 视频生成的演进

从电影化叙事到更可控的视觉创作，Wan 3.0、Seedance 2.5 和 Seedance 2.0 代表了 AI 视频生成的不同思路。通过同一个视觉构思，比较各模型在运动、一致性和创意表达上的表现。

Wan 3.0：电影化视频创作的进阶控制

Wan 3.0 带来更强的视觉理解和生成控制，角色、环境和复杂场景更一致。视频展示了更连贯的动作、更丰富的视听细节和更精确的创作方向，适合高质量视频制作。

Seedance 2.5：更精细的视频动态，创意更强

Seedance 2.5 提升了参考理解，运动更流畅，场景连贯性更强。它的运镜更有表现力，视觉叙事更细腻，在创意探索与可投产效果之间架起桥梁。

Seedance 2.0：可靠的运动与一致的视觉叙事

Seedance 2.0 注重稳定的视频生成，提示词贴合度高、场景转场流畅。它能保持清晰的主体身份和自然的动作，是制作吸引人的视觉叙事的均衡之选。

用 Wan3.0 API 实现更长、更丰富、更可控的视频创意

用 Wan3.0 API 构建更长的叙事片段

一部短片可以跨越多个动作、地点、镜头变化和故事节点展开，而不只是一个孤立的瞬间。Wan3.0 API 很适合城市追逐、微缩冒险、超现实喜剧或电影化动作场景这类片段，故事推进时角色、道具、环境和叙事逻辑都能保持连贯。

用 Wan 3.0 API 制作基于参考的角色视频

当多个角色、服装、环境或视觉参考需要在整个场景中保持可辨认时，Wan 3.0 API 可以把这些素材作为持续的创作指引。适合时尚短片、舞蹈编排、角色展示、历史场景等对外观、造型、空间关系和动作连贯都有要求的制作。

用 Wan3.0-Video API 改造现有素材

现有视频可以作为可控修改的起点，而不必从头重新生成。Wan3.0-Video API 支持替换主体、更换环境、去掉配饰、转换视觉风格、调整部分场景元素或延长现有片段，同时保留应该不变的动作和结构。

用 Wan 3.0 Video API 制作视听内容

围绕对白、音乐、环境音、节奏动作或氛围打造的视频，从一开始就把音频当作创作方向的一部分会更好。Wan 3.0 Video API 适合表演视频、电影感广告、沉浸式故事场景和节奏驱动的动态作品，让声音、画面节奏、角色表演和场景动态紧密同步。

为什么选择 ModelFold，让 Wan3.0 API 工作流更实用

一个模型覆盖文本、帧、参考、文件和链接

wan/3-0-video 走 POST https://modelfold.ai/v1/videos/generations，覆盖文生视频、首帧和首尾帧生视频、参考生视频，以及基于文档或公开网页生成视频。输出分辨率 480P 到 1080P、时长 2 到 30 秒，默认带音频；完整参数见本页的 Playground 和 API 区。各分辨率价格见上方的定价说明和 API 区的费率表。

每次请求只选一种输入模式

首尾帧不能和参考素材、文档或网页一起用，且尾帧必须搭配首帧。文档和网页不能同时使用。参考模式下，在提示词里按传入顺序用 Image1、Video1 或 Audio1 指代素材。首帧和参考图不能是带透明通道的 PNG。

参考视频计入时长和计费

计费秒数等于输出时长加上参考视频总时长，单价不变；带参考视频时，输入加输出时长不能超过 30 秒。时长选自动（-1）时按 30 秒上限收费。ModelFold 新用户可获得 80 免费积分用于测试。

Wan3.0 API 常见问题

Wan3.0 是什么？⌄

Wan3.0 是阿里巴巴的全能视频生成模型，面向更长的视频创作、多模态参考理解、更强的一致性、精确编辑、沉浸式视听生成和电影化控制。它支持多种创作方式，包括文本驱动、图片驱动、基于参考和基于现有视频的流程。

ModelFold 现在能用 Wan3.0 API 吗？⌄

能。Wan3.0 API 已在 ModelFold 上线，可以通过 ModelFold 的 API 流程调用 Wan3.0-Video API 生成视频，并把能力接入你的应用。最新支持的功能和用法，请参阅 ModelFold 上的 Wan3.0-Video API 文档。

Wan 3.0 API 支持哪些视频生成方式？⌄

Wan 3.0 API 支持文生视频、图生视频、首尾帧生视频和基于参考的视频创作。更广泛的多模态能力也让它适合在同一次生成中组合不同创作参考的流程。

Wan3.0-Video API 能理解哪些参考？⌄

Wan3.0-Video API 支持传统文字提示词之外更丰富的创作上下文，可以用不同类型的参考来引导视频生成。视可用的输入选项而定，图片、视频、音频、关键帧、文档和网页等参考素材都能帮助视觉一致性、场景结构、动作设计和整体创作方向。最新支持的参考类型和输入要求，请参阅 ModelFold 上的 Wan3.0-Video API 文档。

Wan3.0 API 能生成最长 30 秒的视频吗？⌄

能。Wan3.0 API 支持原生生成最长 30 秒的视频，让场景有更多空间展开动作、推进镜头、表现角色互动和完整的叙事结构，而不必完全依赖多个短片段拼接。

Wan 3.0 API 如何保持参考一致性？⌄

Wan 3.0 API 的设计目标是在场景和动作变化中保留参考素材的重要细节。随着生成片段推进，角色外观、服装、道具、环境、音频特征和空间关系都能与参考保持更紧密的一致。

Wan3.0-Video API 能编辑现有视频吗？⌄

能。Wan3.0-Video API 支持基于指令和参考引导的视频编辑。可以更换主体、环境、视觉风格、光影等选定元素，同时保留原视频中应保持不变的部分。也支持视频延长，用来续写现有片段。

Wan 3.0 Video API 支持音画同步创作吗？⌄

支持。Wan 3.0 Video API 面向更丰富的视听生成，包括自然的声音表达、语音特征、音乐与节奏、环境音，以及更紧密的音画节奏同步，让生成的视频作为统一的视听体验更完整、更有表现力。

费率表

480p	52 积分 / 5 秒
720p	104 积分 / 5 秒
1080p	208 积分 / 5 秒

按秒计费 —— 上表为各分辨率下「5 秒」的价；时长更长则按比例增加。Playground 与 API 的实际扣费同此口径。

带参考视频时，计费秒数 = 输出秒数 + 参考视频总秒数，单价不变（参考视频总长不超过 15 秒，每条向上取整到整秒；输入 + 输出不超过 30 秒）。时长选 -1（模型自选）按 30 秒收。

快速开始

curl https://modelfold.ai/v1/videos/generations \
  -H "Authorization: Bearer $MODELFOLD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"wan/3-0-video","prompt":"...","resolution":"720p","duration":5}'

wan/3-0-video

在线试跑

更多文生视频模型

Wan 的更多模型

最新热门模型

Wan3.0 API 重新定义多模态视频创作，控制力更强

用 Wan3.0 API 从更丰富的多模态输入开始视频创作

把文字提示词变成完整的视频构思

用图片参考把视觉特征带进动态画面

用视频参考引导动作和场景流动

让声音方向与画面生成联动

用关键帧让场景推进保持一致

把文档和网页带进创作流程

了解 Wan3.0 API 为何是视频创作的一次大飞跃

Wan3.0 API 全能多模态创作

Wan 3.0 API 原生 30 秒叙事

Wan3.0-Video API 像素级参考一致性

Wan 3.0 Video API 精确视频编辑

Wan3.0 API 沉浸式视听生成

Wan3.0-Video API 电影级画面与动作控制

如何在 ModelFold 上用 Wan3.0-Video API 开始创作视频

设置 ModelFold 账号，开通 Wan3.0-Video API

为 Wan3.0-Video API 请求准备创作素材

提交 Wan3.0-Video API 任务并获取结果

通过 Wan 与 Seedance 探索 AI 视频生成的演进

Wan 3.0：电影化视频创作的进阶控制

Seedance 2.5：更精细的视频动态，创意更强

Seedance 2.0：可靠的运动与一致的视觉叙事

用 Wan3.0 API 实现更长、更丰富、更可控的视频创意

用 Wan3.0 API 构建更长的叙事片段

用 Wan 3.0 API 制作基于参考的角色视频

用 Wan3.0-Video API 改造现有素材

用 Wan 3.0 Video API 制作视听内容

为什么选择 ModelFold，让 Wan3.0 API 工作流更实用

一个模型覆盖文本、帧、参考、文件和链接

每次请求只选一种输入模式

参考视频计入时长和计费

Wan3.0 API 常见问题