Wan 3.0 Video
wan / 3-0-video

万相3.0是阿里新一代 All-in-One 视频生成模型,专为多模态创作打造。它支持文本、图像、视频、音频及结构化信息参考,可生成一致性更强、更逼真且更可控的视频内容,全面提升叙事与编辑能力。

Pricing: 8 credits/s for 480P ($0.04/s), 16 credits/s for 720P ($0.08/s), and 32 credits/s for 1080P ($0.16/s), all 20% below official pricing. Note: The billing rule is (input video duration + output video duration) × unit price. High-tier top-ups (+10% bonus) provide an additional 10% reduction in effective pricing.
Input
Start Frame
End Frame

Click to upload or drag and drop images. Single side: [240, 8000] px; aspect ratio ≤ 8:1

Images: JPEG/JPG/PNG/WebP/BMP, Max 20MB

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

0/20000

Click to upload or drag and drop

Supported formats: JPEG, PNG, WEBP, BMP Maximum file size: 20MB; Maximum files: 10

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.

Click to upload or drag and drop

Supported formats: MP4, QUICKTIME Maximum file size: 100MB; Maximum files: 5

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.

Click to upload or drag and drop

Supported formats: MPEG, WAV, X-WAV Maximum file size: 15MB; Maximum files: 5

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.

Text
0 / 1
No items yet. Click Add to start.

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

Text
0 / 1
No items yet. Click Add to start.

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.

Preset resolution levels for video generation.

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

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.

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

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

A configurable parameter. Defaults to true in the Playground.

There is no guarantee that everything can be filtered out; if you are not satisfied with the results, you will need to make your own arrangements.
Output
output typevideo
Examples

Explore different use cases and parameter configurations

README

Complete guide to using wan/3-0-video

万相3.0 API 以更强控制力重塑多模态视频创作

Original image

A short film can develop across multiple actions, locations, camera changes, and story beats without being reduced to a single isolated moment. Wan3.0 API is well suited to sequences such as an urban chase, miniature adventure, surreal comedy, or cinematic action scene where characters, props, environments, and narrative logic need to remain connected as the story progresses.

When several characters, costumes, environments, or visual references need to remain recognizable throughout a scene, Wan 3.0 API can use those materials as persistent creative guidance. This makes it useful for fashion films, choreography sequences, character showcases, historical scenes, and other productions where appearance, styling, spatial relationships, and movement continuity all matter.

Existing video can become the starting point for controlled creative changes rather than being regenerated from scratch. Wan3.0-Video API can support tasks such as replacing a subject, changing an environment, removing an accessory, shifting the visual style, adjusting selected scene elements, or extending an existing sequence while preserving the motion and structure that should remain unchanged.

Videos built around dialogue, music, environmental sound, rhythmic motion, or expressive atmosphere can benefit from treating audio as part of the creative direction from the beginning. Wan 3.0 Video API can support content such as performance videos, cinematic commercials, immersive story scenes, and rhythm-driven motion pieces where sound, visual pacing, character performance, and scene dynamics need to feel closely synchronized.

wan/3-0-video 通过 POST /api/v1/jobs/createTask 调用,覆盖文生视频、首帧及首尾帧生视频、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;完整参数见上方的 Playground 和 API 文档。各分辨率价格见上方价格卡。

首尾帧不能与参考素材、文档或网页同时使用,尾帧必须搭配首帧。文档和网页不能同时使用。参考模式下,请在提示词中按传入顺序以 Image1、Video1 或 Audio1 指代素材。首尾帧图片和参考图片不能是带透明通道的 PNG。

计费时长等于输出时长加参考视频总时长;使用参考视频时,输入与输出时长合计不得超过 30 秒。使用自动时长时,按模型选择的时长计费。Kie.ai 新用户可获得 80 个免费测试积分。

万相3.0 是阿里云推出的全能视频生成模型,专为更长的视频生成、多模态参考理解、更强的一致性、精准编辑、沉浸式音视频生成以及电影级控制而设计。它支持多种创作方式,包括文生视频、图生视频、参考生视频以及基于已有视频的工作流。

是的。在 API Key 页面(kie.ai/api-key)创建 API Key,并在请求头中以 Authorization: Bearer YOUR_API_KEY 发送;然后调用 POST https://api.kie.ai/api/v1/jobs/createTask,model 设为 wan/3-0-video,input 中填写提示词和素材。请求会返回 taskId:可传入 callBackUrl,在视频成功或失败时接收通知,或轮询 GET /api/v1/jobs/recordInfo?taskId=…。全部参数见 万相3.0 视频 API 文档。

万相3.0 视频按秒计费:480P 每秒 8 积分($0.04),720P 每秒 16 积分($0.08),1080P 每秒 32 积分($0.16),因此 5 秒 1080P 视频为 160 积分($0.80)。计费时长 = 输出时长 + 参考视频总时长;duration 设为 -1 时,按模型自动选择的时长计费。价格可能会调整,请以本页顶部价格卡显示的价格为准。

没有。Kie.ai 没有订阅或无限量套餐:积分按量付费、永不过期,每条 万相3.0 视频按上述价格按秒计费。默认情况下,每个账号每 10 秒最多可提交 20 个新的生成请求,通常可同时运行 100 个以上任务;超出限制的请求返回 429 代码,且不会进入队列。

万相3.0 API 支持文生视频、图生视频、首尾帧视频生成以及参考生视频。其更广泛的多模态能力,也使其非常适合在同一生成过程中结合不同创意参考素材的工作流。

参考模式下,wan/3-0-video 最多接收 10 张图片(JPEG、不含透明通道的 PNG、BMP 或 WebP,单边 240–8000 像素,最大 20MB)、5 段视频(MP4 或 MOV,每段 1–15 秒、合计不超过 15 秒,每段最大 100MB)和 5 段音频(WAV 或 MP3,每段 1–15 秒、合计不超过 15 秒,最大 15MB),在提示词中按传入顺序以 Image1、Video1、Audio1 指代。也可以从首帧(可加尾帧)、一个文档(如 DOCX、XLSX、PPTX、PDF、TXT 或 MD,最大 100MB、50 页)或一个公开网页开始生成。首尾帧不能与参考素材、文档或网页同时使用,文档和网页也不能同时使用。

是的。Wan3.0 API 原生支持长达30秒的视频生成,为动作延展、运镜推进、角色互动以及完整的叙事结构提供了更大的创作空间,而无需完全依赖多个短片段。

Wan 3.0 API 旨在在场景和动态变化中保留所提供参考素材的重要细节。随着生成序列的推进,人物外观、服饰、道具、环境、音频特征及空间关系能够保持更紧密的对齐。

支持。万相3.0视频 API 提供基于指令和参考素材的视频编辑功能。您可以在保留原视频特定部分不变的基础上,灵活修改视频主体、环境、视觉风格、光影等元素。此外,也支持视频续写功能,为您自然延续现有视频画面。编辑或续写时,将视频传入 reference_video_urls,并在提示词中以 Video1 指代、写明修改或续写内容(阿里云万相3.0 API 参考文档的做法);参考视频时长与输出时长一起计费。

支持。万相3.0视频 API 致力于提供更丰富的音视频生成体验,包括自然的声音表达、语音特征、音乐与节奏以及环境音效,并能实现声音与画面节奏的精准同步。这将为您带来视听高度统一的沉浸式体验,让生成的视频内容更生动完整、极具感染力。