# Doubao Seedance 2.0 系列教程 > 全部文档的索引见 [llms.txt](https://ark.volcengine.com/llms.txt)；在任一文档页面地址后追加 `.md` 可获取该页的 Markdown 版本。 Doubao Seedance 2.0 系列（下文简称 Seedance 2.0 系列）模型支持图像、视频、音频、文本等多种模态内容输入，具备视频生成、视频编辑、视频延长等能力，可高精度还原物品细节、音色、效果、风格、运镜等，保持稳定角色特征，赋予使用者如同导演般的掌控权。本文介绍 Seedance 2.0 系列模型的专属能力，帮助您快速实现 [Video Generation API](/docs/create-video-generation-task-api) 调用。 > **说明：模型开通条件** > > 开通 Seedance 2.0 系列模型前，请确保您满足以下任一条件： > * **【推荐】账户余额 > 200 元**（[前往充值](https://console.volcengine.com/finance/fund/recharge)） > * **【推荐】购买 200 元档位及以上专属节省计划**，购买入口：[购买节省计划](https://console.volcengine.com/common-buy/AI-SavingsPlans%7C%7Cd682ppeeq1mp7kd5q0e0)。 > * 已购买 Seedance 2.0 系列资源包且有可用余量 （[前往购买](https://console.volcengine.com/common-buy/fast/ark_bd%7C%7Cd682ppeeq1mp7kd5q0e0)） > > 详细规则见 [Seedance 2.5 与 Seedance 2.0 系列模型开通、使用与退订说明](/docs/seedance-model-activation-usage-and-refund)。 # 新手入门 本入门教程专为 **API 新手用户** 设计，帮助您一键搭建 Python 开发环境、完成虚拟环境创建和方舟 SDK 安装，并提供直接可运行的 Seedance 2.0 系列调用代码，您只需修改对应的输入素材，即可开始您的视频生成创作。 **1. 准备工作** 在开始之前，请确保您已经完成以下准备： 1. **注册账号**：确保您拥有火山引擎账号并已 [登录](https://console.volcengine.com/)。 2. **获取 API Key**：访问 [API Key 管理页面](https://ark.volcengine.com/region:cn-beijing/apikey)，点击 **创建 API Key**，并复制保存您的 API Key。注意请妥善保管您的 API Key，不要泄露给他人。 3. [**开通模型**](https://ark.volcengine.com/region:cn-beijing/openManagement?LLM=%7B%7D&advancedActiveKey=model&projectName=default&tab=ComputerVision)：请确保您的账户余额大于等于 200 元，或已 [购买资源包](https://console.volcengine.com/common-buy/fast/ark_bd%7C%7Cd682ppeeq1mp7kd5q0e0)，否则无法开通 Seedance 2.0 系列模型。 4. **下载并解压文件**：点击下载下方附件，将其解压到您的本地目录（如桌面或“下载”文件夹）。 [ark_seedance2.0_quickstart_package.zip](https://arkdocs.tos-cn-beijing.volces.com/files/video-generation/ark_seedance2.0_quickstart_package_arkruntime-20260909.zip) **2.操作步骤** **选项：Windows 用户** 1. 进入 `scripts/init_dev_env` 目录。 2. 双击运行 `setup_windows.bat`。 3. 脚本会自动执行以下操作： * 下载 uv 工具。 * 自动下载 Python 3.12（如果不干扰您的系统 Python）。 * 创建虚拟环境 .`venv`。 * 安装方舟 SDK。 4. 完成后，在项目根目录会生成一个 `run_demo.bat`。 5. 双击 `run_demo.bat`，即可运行 Python SDK 示例代码(python/demo_standard.py)。 **选项：macOS 用户** 1. 打开终端，进入 `scripts/init_dev_env` 目录。 2. 运行构建脚本： ```Plain Text ./setup_mac.sh ``` 3. 脚本会自动配置好所有环境。 4. 完成后，在项目根目录会生成一个 `run_demo.sh`。 5. 运行 `./run_demo.sh` 即可运行 Python SDK 示例代码(python/demo_standard.py)。 **3.运行说明** 运行脚本后，您将看到如下流程： 1. **API Key 校验**：脚本会自动检测您本地是否配置了`ARK_API_KEY`环境变量。如果没有，会提示您手动输入。 2. **素材预览**：脚本会自动在您的默认浏览器中弹出一个本地生成的 HTML 页面，直观地展示本次任务的文本提示词、待替换的参考图片以及原始参考视频。 3. **任务创建与轮询**：脚本向火山方舟服务器发起异步请求。由于视频生成需要一定时间，控制台会每隔 30 秒打印一次任务状态（如 `running`等）。 4. **获取结果**：任务成功后，控制台会输出一段最终生成的视频 URL。您可以复制该链接到浏览器下载或在线播放。 **4.下一步** 在成功跑通本示例后，您可以尝试修改 `python/``demo_standard.py`，来打造您专属的视频生成任务： 1. 修改文本提示词 找到代码中的 `user_content` 变量，更改为您想要的画面描述。 2. 替换输入素材 (图片、视频、音频) 您可以将 `reference_image_url`、`reference_video_url` 和 `reference_audio_url` 替换为您自己的素材链接。 **注意**：请确保 URL 是公网可公开访问的链接（建议存放在 TOS 对象存储服务中，并配置为公共读）。 3. 继续学习下文中丰富的使用示例。 # 模型能力 Seedance 2.0 系列模型目前包括 Doubao Seedance 2.0（下文简称 Seedance 2.0）、Doubao Seedance 2.0 Fast（下文简称 Seedance 2.0 Fast）和 Doubao Seedance 2.0 Mini（下文简称 Seedance 2.0 Mini）。三者支持的功能基本一致，主要区别在于生成品质与成本的取舍： - 追求最高生成品质，推荐使用 Seedance 2.0； - 兼顾成本与生成速度，不要求极致品质，推荐使用 Seedance 2.0 Fast； - 追求更低成本，推荐使用 Seedance 2.0 Mini。 |模型名称 | |[Seedance 2.0](https://ark.volcengine.com/region:cn-beijing/model/detail?Id=doubao-seedance-2-0&projectName=default) |[Seedance 2.0 Fast](https://ark.volcengine.com/region:cn-beijing/model/detail?Id=doubao-seedance-2-0-fast&projectName=default) |[Seedance 2.0 Mini](https://ark.volcengine.com/region:cn-beijing/model/detail?Id=doubao-seedance-2-0-mini&projectName=default) | |---|---|---|---|---| |Model ID | |doubao-seedance-2-0-260128 |doubao-seedance-2-0-fast-260128 |doubao-seedance-2-0-mini-260615 | \ | | | | | | |[文生视频](/docs/video-generation-tutorial#4e74bcee) | |✓ |✓ |✓ | |[图生视频-首帧](/docs/video-generation-tutorial#979b2d28) | |✓ |✓ |✓ | |[图生视频-首尾帧](/docs/video-generation-tutorial#0d55ca07) | |✓ |✓ |✓ | |[全模态参考](/docs/seedance-2-0#50e1b4ea)【New】 |图片参考 |✓ |✓ |✓ | |^^| | | | | \ | |视频参考 |✓ |✓ |✓ | |^^| | | | | \ | |组合参考 |✓ |✓ |✓ | \ | | | | | | \ | |* 图片 + 音频 | | | | \ | |* 图片 + 视频 | | | | \ | |* 视频 + 音频 | | | | \ | |* 图片 + 视频 + 音频 | | | | |[编辑视频](/docs/seedance-2-0#75a28782)【New】 | |✓ |✓ |✓ | |[延长视频](/docs/seedance-2-0#46d77653)【New】 | |✓ |✓ |✓ | |[生成有声视频](/docs/video-generation-tutorial#979b2d28) | |✓ |✓ |✓ | \ |> "generate_audio": "true" | | | | | |[联网搜索工具](/docs/seedance-2-0#c40ed3ef)【New】 | |✓ |✓ |✓ | |[样片模式](/docs/video-generation-tutorial#5acd28c8) | |✗ |✗ |✗ | |[返回视频产物对应的尾帧图](/docs/video-generation-tutorial#141cf7fa) | |✓ |✓ |✓ | \ |> "return_last_frame": | | | | | \ |> "true" | | | | | |[输出视频规格](/docs/video-generation-tutorial#9fe4cce0) |输出分辨率 |* 480p（8bit 位深） |* 480p（8bit 位深） |* 480p（8bit 位深） | \ | | |* 720p（8bit 位深） |* 720p（8bit 位深） |* 720p（8bit 位深） | \ | | |* 1080p（8bit 位深） | | | \ | | |* 4k（10bit 位深） | | | \ | |> "resolution": "720p" | | | | | |输出宽高比 |21:9, 16:9, 4:3, |21:9, 16:9, 4:3, |21:9, 16:9, 4:3, | \ | |> "ratio":"16:9" | 1:1, 3:4, 9:16 | 1:1, 3:4, 9:16 | 1:1, 3:4, 9:16 | | |输出时长 |4~15 秒 |4~15 秒 |4~15 秒 | \ | |> "duration": 5 | | | | | |输出视频格式 |mp4 |mp4 |mp4 | |[离线推理](/docs/video-generation-tutorial#c3588bd1) | |✗ |✗ |✗ | \ |> "service_tier": "flex" | | | | | |在线推理限流 |最大 RPM |非 4k 分辨率： |* 企业用户：600 |* 企业用户：600 | \ | | |* 企业用户：600 |* 个人用户：180 |* 个人用户：180 | \ | | |* 个人用户：180 | | | \ | | | | | | \ | | |4k 分辨率： | | | \ | | |* 企业用户：15 | | | \ | | |* 个人用户：15 | | | | |最大并发数 |非 4k 分辨率： |* 企业用户：10 |* 企业用户：10 | \ | | |* 企业用户：10 |* 个人用户：3 |* 个人用户：3 | \ | | |* 个人用户：3 | | | \ | | | | | | \ | | |4k 分辨率： | | | \ | | |* 企业用户：1 | | | \ | | |* 个人用户：1 | | | |离线推理限流 |TPD |\- |\- |\- | # 基础使用 ## 全模态参考 输入文本、参考图、视频（可带音轨）和音频等内容，来生成一段新视频。可继承参考图片的角色形象、视觉风格、画面构图；参考视频的主体内容、运镜方式、动作表现、整体风格；以及参考音频的音色、音乐旋律、对话内容等核心信息。 效果预览如下（访问 [模型卡片](https://ark.volcengine.com/region:cn-beijing/model/detail?Id=doubao-seedance-2-0) 查看更多示例）： |输入：文本 |输入：图片、视频、音频 |输出 | |---|---|---| |全程使用**视频1**的第一视角构图，全程使用**音频1**作为背景音乐。第一人称视角果茶宣传广告，seedance牌「苹苹安安」苹果果茶限定款；首帧为**图片1**，你的手摘下一颗带晨露的阿克苏红苹果，清脆的苹果碰撞声；2-4 秒：快速切镜，你的手将苹果块投入雪克杯，加入冰块与茶底，用力摇晃，冰块碰撞声与摇晃声卡点轻快鼓点，背景音：「鲜切现摇」；4-6 秒：第一人称成品特写，分层果茶倒入透明杯，你的手轻挤奶盖在顶部铺展，在杯身贴上粉红包标，镜头拉近看奶盖与果茶的分层纹理；6-8 秒：第一人称手持举杯，你将**图片2**中的果茶举到镜头前（模拟递到观众面前的视角），杯身标签清晰可见，背景音「来一口鲜爽」，尾帧定格为**图片2**。背景声音统一为女生音色。 |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/0ba05cd435f543c5bc65c378d94d094a) |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/dab46ce2289a4a8ead76711bb02f2e1d) | \ | | | | \ | | | | \ | |![](https://p9-arcosite.byteimg.com/tos-cn-i-goo7wpa0wc/37ef4b6af8944a6d9b54ef1c541c1b0e~tplv-goo7wpa0wc-image.image) | | \ | | | | \ | |![](https://p9-arcosite.byteimg.com/tos-cn-i-goo7wpa0wc/7b904d6b46d24f059de7697620058b7f~tplv-goo7wpa0wc-image.image) | | \ | | | | \ | | | | **选项：Python** ```Python import os import time # Install SDK: python -m pip install --upgrade arkruntime from arkruntime import Ark client = Ark( # The base URL for model invocation base_url='https://ark.cn-beijing.volces.com/api/v3', # Get API Key：https://ark.volcengine.com/region:cn-beijing/apikey api_key=os.environ.get("ARK_API_KEY"), ) if __name__ == "__main__": print("----- create request -----") create_result = client.content_generation.tasks.create( model="doubao-seedance-2-0-260128", # Replace with Model ID content=[ { "type": "text", "text": "全程使用视频1的第一视角构图，全程使用音频1作为背景音乐。第一人称视角果茶宣传广告，seedance牌「苹苹安安」苹果果茶限定款；首帧为图片1，你的手摘下一颗带晨露的阿克苏红苹果，轻脆的苹果碰撞声；2-4 秒：快速切镜，你的手将苹果块投入雪克杯，加入冰块与茶底，用力摇晃，冰块碰撞声与摇晃声卡点轻快鼓点，背景音：「鲜切现摇」；4-6 秒：第一人称成品特写，分层果茶倒入透明杯，你的手轻挤奶盖在顶部铺展，在杯身贴上粉红包标，镜头拉近看奶盖与果茶的分层纹理；6-8 秒：第一人称手持举杯，你将图片2中的果茶举到镜头前（模拟递到观众面前的视角），杯身标签清晰可见，背景音「来一口鲜爽」，尾帧定格为图片2。背景声音统一为女生音色。", }, { "type": "image_url", "image_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_tea_pic1-zh.jpg" }, "role": "reference_image", }, { "type": "image_url", "image_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_tea_pic2-zh.jpg" }, "role": "reference_image", }, { "type": "video_url", "video_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_tea_video1.mp4" }, "role": "reference_video", }, { "type": "audio_url", "audio_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/files/video-generation/r2v_tea_audio1.mp3" }, "role": "reference_audio", }, ], generate_audio=True, ratio="16:9", duration=11, watermark=True, ) print(create_result) # Polling query section print("----- polling task status -----") task_id = create_result.id while True: get_result = client.content_generation.tasks.get(task_id=task_id) status = get_result.status if status == "succeeded": print("----- task succeeded -----") print(get_result) break elif status == "failed": print("----- task failed -----") print(f"Error: {get_result.error}") break else: print(f"Current status: {status}, Retrying after 30 seconds...") time.sleep(30) ``` **选项：Java** ```Java package com.ark.sample; import com.volcengine.ark.runtime.models.content_generation.*; import com.volcengine.ark.runtime.service.ArkService; import okhttp3.ConnectionPool; import okhttp3.Dispatcher; import java.util.ArrayList; import java.util.List; import java.util.concurrent.TimeUnit; public class ContentGenerationTaskExample { // Client initialization static String apiKey = System.getenv("ARK_API_KEY"); static ConnectionPool connectionPool = new ConnectionPool(5, 1, TimeUnit.SECONDS); static Dispatcher dispatcher = new Dispatcher(); static ArkService service = ArkService.builder() .baseUrl("https://ark.cn-beijing.volces.com/api/v3") // The base URL for model invocation .dispatcher(dispatcher) .connectionPool(connectionPool) .apiKey(apiKey) .build(); public static void main(String[] args) { // Model ID final String modelId = "doubao-seedance-2-0-260128"; // Text prompt final String prompt = "全程使用视频1的第一视角构图，全程使用音频1作为背景音乐。第一人称视角果茶宣传广告，seedance牌「苹苹安安」苹果果茶限定款；" + "首帧为图片1，你的手摘下一颗带晨露的阿克苏红苹果，轻脆的苹果碰撞声；" + "2-4 秒：快速切镜，你的手将苹果块投入雪克杯，加入冰块与茶底，用力摇晃，冰块碰撞声与摇晃声卡点轻快鼓点，背景音：「鲜切现摇」；" + "4-6 秒：第一人称成品特写，分层果茶倒入透明杯，你的手轻挤奶盖在顶部铺展，在杯身贴上粉红包标，镜头拉近看奶盖与果茶的分层纹理；" + "6-8 秒：第一人称手持举杯，你将图片2中的果茶举到镜头前（模拟递到观众面前的视角），杯身标签清晰可见，背景音「来一口鲜爽」，尾帧定格为图片2。" + "背景声音统一为女生音色。"; // Example resource URLs final String refImage1 = "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_tea_pic1-zh.jpg"; final String refImage2 = "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_tea_pic2-zh.jpg"; final String refVideo = "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_tea_video1.mp4"; final String refAudio = "https://arkdocs.tos-cn-beijing.volces.com/files/video-generation/r2v_tea_audio1.mp3"; // Output video parameters final boolean generateAudio = true; final String videoRatio = "16:9"; final long videoDuration = 11L; final boolean showWatermark = true; System.out.println("----- create request -----"); // Build request content List contents = new ArrayList<>(); // 1. Text prompt contents.add(ContentItem.builder() .type(ContentType.TEXT) .text(prompt) .build()); // 2. Reference image 1 contents.add(ContentItem.builder() .type(ContentType.IMAGE_URL) .imageUrl(ImageURL.builder() .url(refImage1) .build()) .role("reference_image") .build()); // 3. Reference image 2 contents.add(ContentItem.builder() .type(ContentType.IMAGE_URL) .imageUrl(ImageURL.builder() .url(refImage2) .build()) .role("reference_image") .build()); // 4. Reference video contents.add(ContentItem.builder() .type(ContentType.VIDEO_URL) .videoUrl(VideoURL.builder() .url(refVideo) .build()) .role("reference_video") .build()); // 5. Reference audio contents.add(ContentItem.builder() .type(ContentType.AUDIO_URL) .audioUrl(AudioURL.builder() .url(refAudio) .build()) .role("reference_audio") .build()); // Create video generation task CreateContentGenerationTaskRequest createRequest = CreateContentGenerationTaskRequest.builder() .generateAudio(generateAudio) .model(modelId) .content(contents) .ratio(videoRatio) .duration(videoDuration) .watermark(showWatermark) .build(); CreateContentGenerationTaskResponse createResult = service.createContentGenerationTask(createRequest); System.out.println("Task Created: " + createResult); // Get task details and poll status String taskId = createResult.getId(); pollTaskStatus(taskId); } /** * Poll task status * @param taskId Task ID */ private static void pollTaskStatus(String taskId) { String getRequest = taskId; System.out.println("----- polling task status -----"); try { while (true) { ContentGenerationTask getResponse = service.getContentGenerationTask(getRequest); String status = getResponse.getStatus().toString(); if ("succeeded".equalsIgnoreCase(status)) { System.out.println("----- task succeeded -----"); System.out.println(getResponse); break; } else if ("failed".equalsIgnoreCase(status)) { System.out.println("----- task failed -----"); if (getResponse.getError() != null) { System.out.println("Error: " + getResponse.getError().getMessage()); } break; } else { System.out.printf("Current status: %s, Retrying in 10 seconds...%n", status); TimeUnit.SECONDS.sleep(10); } } } catch (InterruptedException ie) { Thread.currentThread().interrupt(); System.err.println("Polling interrupted"); } catch (Exception e) { System.err.println("Error occurred: " + e.getMessage()); } finally { service.shutdownExecutor(); } } } ``` **选项：Go** ```Go package main import ( "context" "fmt" "os" "time" "github.com/volcengine/ark-runtime-go/arkruntime" model "github.com/volcengine/ark-runtime-go/arkruntime/model/contentgeneration" ) func main() { // Initialize Ark client client := arkruntime.NewClientWithApiKey( os.Getenv("ARK_API_KEY"), // The base URL for model invocation arkruntime.WithBaseUrl("https://ark.cn-beijing.volces.com/api/v3"), ) ctx := context.Background() // Model ID modelID := "doubao-seedance-2-0-260128" // Text prompt prompt := "全程使用视频1的第一视角构图，全程使用音频1作为背景音乐。第一人称视角果茶宣传广告，seedance牌「苹苹安安」苹果果茶限定款；" + "首帧为图片1，你的手摘下一颗带晨露的阿克苏红苹果，轻脆的苹果碰撞声；" + "2-4 秒：快速切镜，你的手将苹果块投入雪克杯，加入冰块与茶底，用力摇晃，冰块碰撞声与摇晃声卡点轻快鼓点，背景音：「鲜切现摇」；" + "4-6 秒：第一人称成品特写，分层果茶倒入透明杯，你的手轻挤奶盖在顶部铺展，在杯身贴上粉红包标，镜头拉近看奶盖与果茶的分层纹理；" + "6-8 秒：第一人称手持举杯，你将图片2中的果茶举到镜头前（模拟递到观众面前的视角），杯身标签清晰可见，背景音「来一口鲜爽」，尾帧定格为图片2。" + "背景声音统一为女生音色。" // Example resource URLs refImage1 := "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_tea_pic1-zh.jpg" refImage2 := "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_tea_pic2-zh.jpg" refVideo := "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_tea_video1.mp4" refAudio := "https://arkdocs.tos-cn-beijing.volces.com/files/video-generation/r2v_tea_audio1.mp3" // Output video parameters generateAudio := true videoRatio := "16:9" videoDuration := int64(11) showWatermark := true // 1. Create video generation task fmt.Println("----- create request -----") createReq := &model.CreateContentGenerationTaskRequest{ Model: modelID, GenerateAudio: model.NewOptBool(generateAudio), Ratio: model.NewOptString(videoRatio), Duration: model.NewOptInt64(videoDuration), Watermark: model.NewOptBool(showWatermark), Content: []model.ContentItem{ { Type: model.ContentTypeText, Text: model.NewOptString(prompt), }, { Type: model.ContentTypeImageURL, ImageURL: model.NewOptImageURL(model.ImageURL{ URL: refImage1, }), Role: model.NewOptString("reference_image"), }, { Type: model.ContentTypeImageURL, ImageURL: model.NewOptImageURL(model.ImageURL{ URL: refImage2, }), Role: model.NewOptString("reference_image"), }, { Type: model.ContentTypeVideoURL, VideoURL: model.NewOptVideoURL(model.VideoURL{ URL: refVideo, }), Role: model.NewOptString("reference_video"), }, { Type: model.ContentTypeAudioURL, AudioURL: model.NewOptAudioURL(model.AudioURL{ URL: refAudio, }), Role: model.NewOptString("reference_audio"), }, }, } createResp, err := client.CreateContentGenerationTask(ctx, createReq) if err != nil { fmt.Printf("create content generation error: %v\n", err) return } taskID := createResp.ID fmt.Printf("Task Created with ID: %s\n", taskID) // 2. Poll task status pollTaskStatus(ctx, client, taskID) } // poll task status func pollTaskStatus(ctx context.Context, client *arkruntime.Client, taskID string) { fmt.Println("----- polling task status -----") for { getReq := taskID getResp, err := client.GetContentGenerationTask(ctx, getReq) if err != nil { fmt.Printf("get content generation task error: %v\n", err) return } status := getResp.Status if status == "succeeded" { fmt.Println("----- task succeeded -----") fmt.Printf("Task ID: %s \n", getResp.ID) fmt.Printf("Model: %s \n", getResp.Model) fmt.Printf("Video URL: %s \n", getResp.Content.Or(model.TaskContent{}).VideoURL.Or("")) fmt.Printf("Completion Tokens: %d \n", getResp.Usage.Or(model.TaskUsage{}).CompletionTokens) fmt.Printf("Created At: %d, Updated At: %d\n", getResp.CreatedAt.Or(0), getResp.UpdatedAt.Or(0)) return } else if status == "failed" { fmt.Println("----- task failed -----") if getResp.Error.IsSet() { fmt.Printf("Error Code: %s, Message: %s\n", getResp.Error.Value.Code, getResp.Error.Value.Message) } return } else { fmt.Printf("Current status: %s, Retrying in 10 seconds... \n", status) time.Sleep(10 * time.Second) } } } ``` > **说明** > > * 您可任意组合以下模态内容，注意不支持“文本+音频”、“纯音频” 输入。 > > * 文本 > > * 图片：0~9 张 > > * 视频：0~3 个 > > * 音频：0~3 个 > * **进阶用法**：全模态生视频可通过提示词指定参考图片作为首帧/尾帧，间接实现“首尾帧+全模态参考”效果。若需严格保障首尾帧和指定图片一致，**优先使用图生视频-首尾帧**（配置 role 为 first_frame/last_frame）。 > * 各个模态信息输入要求参见 [全模态输入](/docs/seedance-2-0#50e1b4ea)。 ## 编辑视频 您可以提供待编辑的视频、参考图片或音频，并结合使用提示词，完成多种视频编辑任务，例如：替换视频主体、视频中对象增删改、局部画面重绘/修复等。 效果预览如下（访问 [模型卡片](https://ark.volcengine.com/region:cn-beijing/model/detail?Id=doubao-seedance-2-0) 查看更多示例）： |输入：文本 |输入：视频&图片 |输出 | |---|---|---| |将**视频1**礼盒中的香水替换成**图像1**中的面霜，运镜不变 |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/0a1afd3250d84b8995e9c0aa61b57d38) |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/fd7bcf4eaf504f50aeeebd48ce35c06a) | \ | | | | \ | | | | \ | | | | \ | |![](https://p9-arcosite.byteimg.com/tos-cn-i-goo7wpa0wc/791b783fc6cd4394b13f41b66b5ff461~tplv-goo7wpa0wc-image.image) | | **选项：Python** ```Python import os import time # Install SDK: python -m pip install --upgrade arkruntime from arkruntime import Ark client = Ark( # The base URL for model invocation base_url='https://ark.cn-beijing.volces.com/api/v3', # Get API Key：https://ark.volcengine.com/region:cn-beijing/apikey api_key=os.environ.get("ARK_API_KEY"), ) if __name__ == "__main__": print("----- create request -----") create_result = client.content_generation.tasks.create( model="doubao-seedance-2-0-260128", # Replace with Model ID content=[ { "type": "text", "text": "将视频1礼盒中的香水替换成图片1中的面霜，运镜不变", }, { "type": "image_url", "image_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_edit_pic1.jpg" }, "role": "reference_image", }, { "type": "video_url", "video_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_edit_video1.mp4" }, "role": "reference_video", }, ], generate_audio=True, ratio="16:9", duration=5, watermark=True, ) print(create_result) # Polling query section print("----- polling task status -----") task_id = create_result.id while True: get_result = client.content_generation.tasks.get(task_id=task_id) status = get_result.status if status == "succeeded": print("----- task succeeded -----") print(get_result) break elif status == "failed": print("----- task failed -----") print(f"Error: {get_result.error}") break else: print(f"Current status: {status}, Retrying after 30 seconds...") time.sleep(30) ``` **选项：Java** ```Java package com.ark.sample; import com.volcengine.ark.runtime.models.content_generation.*; import com.volcengine.ark.runtime.service.ArkService; import okhttp3.ConnectionPool; import okhttp3.Dispatcher; import java.util.ArrayList; import java.util.List; import java.util.concurrent.TimeUnit; public class ContentGenerationTaskExample { // Client initialization static String apiKey = System.getenv("ARK_API_KEY"); static ConnectionPool connectionPool = new ConnectionPool(5, 1, TimeUnit.SECONDS); static Dispatcher dispatcher = new Dispatcher(); static ArkService service = ArkService.builder() .baseUrl("https://ark.cn-beijing.volces.com/api/v3") // The base URL for model invocation .dispatcher(dispatcher) .connectionPool(connectionPool) .apiKey(apiKey) .build(); public static void main(String[] args) { // Model ID final String modelId = "doubao-seedance-2-0-260128"; // Text prompt final String prompt = "将视频1礼盒中的香水替换成图片1中的面霜，运镜不变"; // Example resource URLs final String refImage1 = "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_edit_pic1.jpg"; final String refVideo = "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_edit_video1.mp4"; // Output video parameters final boolean generateAudio = true; final String videoRatio = "16:9"; final long videoDuration = 5L; final boolean showWatermark = true; System.out.println("----- create request -----"); // Build request content List contents = new ArrayList<>(); // 1. Text prompt contents.add(ContentItem.builder() .type(ContentType.TEXT) .text(prompt) .build()); // 2. Reference image 1 contents.add(ContentItem.builder() .type(ContentType.IMAGE_URL) .imageUrl(ImageURL.builder() .url(refImage1) .build()) .role("reference_image") .build()); // 3. Reference video contents.add(ContentItem.builder() .type(ContentType.VIDEO_URL) .videoUrl(VideoURL.builder() .url(refVideo) .build()) .role("reference_video") .build()); // Create video generation task CreateContentGenerationTaskRequest createRequest = CreateContentGenerationTaskRequest.builder() .generateAudio(generateAudio) .model(modelId) .content(contents) .ratio(videoRatio) .duration(videoDuration) .watermark(showWatermark) .build(); CreateContentGenerationTaskResponse createResult = service.createContentGenerationTask(createRequest); System.out.println("Task Created: " + createResult); // Get task details and poll status String taskId = createResult.getId(); pollTaskStatus(taskId); } /** * Poll task status * @param taskId Task ID */ private static void pollTaskStatus(String taskId) { String getRequest = taskId; System.out.println("----- polling task status -----"); try { while (true) { ContentGenerationTask getResponse = service.getContentGenerationTask(getRequest); String status = getResponse.getStatus().toString(); if ("succeeded".equalsIgnoreCase(status)) { System.out.println("----- task succeeded -----"); System.out.println(getResponse); break; } else if ("failed".equalsIgnoreCase(status)) { System.out.println("----- task failed -----"); if (getResponse.getError() != null) { System.out.println("Error: " + getResponse.getError().getMessage()); } break; } else { System.out.printf("Current status: %s, Retrying in 10 seconds...%n", status); TimeUnit.SECONDS.sleep(10); } } } catch (InterruptedException ie) { Thread.currentThread().interrupt(); System.err.println("Polling interrupted"); } catch (Exception e) { System.err.println("Error occurred: " + e.getMessage()); } finally { service.shutdownExecutor(); } } } ``` **选项：Go** ```Go package main import ( "context" "fmt" "os" "time" "github.com/volcengine/ark-runtime-go/arkruntime" model "github.com/volcengine/ark-runtime-go/arkruntime/model/contentgeneration" ) func main() { // Initialize Ark client client := arkruntime.NewClientWithApiKey( os.Getenv("ARK_API_KEY"), // The base URL for model invocation arkruntime.WithBaseUrl("https://ark.cn-beijing.volces.com/api/v3"), ) ctx := context.Background() // Model ID modelID := "doubao-seedance-2-0-260128" // Text prompt prompt := "将视频1礼盒中的香水替换成图片1中的面霜，运镜不变" // Example resource URLs refImage1 := "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/r2v_edit_pic1.jpg" refVideo1 := "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_edit_video1.mp4" // Output video parameters generateAudio := true videoRatio := "16:9" videoDuration := int64(5) showWatermark := true // 1. Create video generation task fmt.Println("----- create request -----") createReq := &model.CreateContentGenerationTaskRequest{ Model: modelID, GenerateAudio: model.NewOptBool(generateAudio), Ratio: model.NewOptString(videoRatio), Duration: model.NewOptInt64(videoDuration), Watermark: model.NewOptBool(showWatermark), Content: []model.ContentItem{ { Type: model.ContentTypeText, Text: model.NewOptString(prompt), }, { Type: model.ContentTypeImageURL, ImageURL: model.NewOptImageURL(model.ImageURL{ URL: refImage1, }), Role: model.NewOptString("reference_image"), }, { Type: model.ContentTypeVideoURL, VideoURL: model.NewOptVideoURL(model.VideoURL{ URL: refVideo1, }), Role: model.NewOptString("reference_video"), }, }, } createResp, err := client.CreateContentGenerationTask(ctx, createReq) if err != nil { fmt.Printf("create content generation error: %v\n", err) return } taskID := createResp.ID fmt.Printf("Task Created with ID: %s\n", taskID) // 2. Poll task status pollTaskStatus(ctx, client, taskID) } // poll task status func pollTaskStatus(ctx context.Context, client *arkruntime.Client, taskID string) { fmt.Println("----- polling task status -----") for { getReq := taskID getResp, err := client.GetContentGenerationTask(ctx, getReq) if err != nil { fmt.Printf("get content generation task error: %v\n", err) return } status := getResp.Status if status == "succeeded" { fmt.Println("----- task succeeded -----") fmt.Printf("Task ID: %s \n", getResp.ID) fmt.Printf("Model: %s \n", getResp.Model) fmt.Printf("Video URL: %s \n", getResp.Content.Or(model.TaskContent{}).VideoURL.Or("")) fmt.Printf("Completion Tokens: %d \n", getResp.Usage.Or(model.TaskUsage{}).CompletionTokens) fmt.Printf("Created At: %d, Updated At: %d\n", getResp.CreatedAt.Or(0), getResp.UpdatedAt.Or(0)) return } else if status == "failed" { fmt.Println("----- task failed -----") if getResp.Error.IsSet() { fmt.Printf("Error Code: %s, Message: %s\n", getResp.Error.Value.Code, getResp.Error.Value.Message) } return } else { fmt.Printf("Current status: %s, Retrying in 10 seconds... \n", status) time.Sleep(10 * time.Second) } } } ``` ## 延长视频 在原有视频基础上，向前或者向后延长视频，或多个视频片段（最多 3 个视频片段）串联成一个连贯视频。 效果预览如下（访问 [模型卡片](https://ark.volcengine.com/region:cn-beijing/model/detail?Id=doubao-seedance-2-0) 查看更多示例）： |输入：文本 |输入：待延长视频 |输出 | |---|---|---| |**视频1**中的拱形窗户打开，进入美术馆室内，接**视频2**，之后镜头进入画内，接**视频3** |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/54519ff7266d4f1caa12b8cc95e2dd1d) |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/849b3f86f609495ca09d559aa14c79ed) | \ | | | | \ | | | | \ | | | | \ | |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/b15d56c80c884faa8526beb6ca540b98) | | \ | | | | \ | | | | \ | | | | \ | |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/f5d327311e094361b15dca0a37b14ab4) | | **选项：Python** ```Python import os import time # Install SDK: python -m pip install --upgrade arkruntime from arkruntime import Ark client = Ark( # The base URL for model invocation base_url='https://ark.cn-beijing.volces.com/api/v3', # Get API Key：https://ark.volcengine.com/region:cn-beijing/apikey api_key=os.environ.get("ARK_API_KEY"), ) if __name__ == "__main__": print("----- create request -----") create_result = client.content_generation.tasks.create( model="doubao-seedance-2-0-260128", # Replace with Model ID content=[ { "type": "text", "text": "视频1中的拱形窗户打开，进入美术馆室内，接视频2，之后镜头进入画内，接视频3", }, { "type": "video_url", "video_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video1.mp4" }, "role": "reference_video", }, { "type": "video_url", "video_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video2.mp4" }, "role": "reference_video", }, { "type": "video_url", "video_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video3.mp4" }, "role": "reference_video", }, ], generate_audio=True, ratio="16:9", duration=8, watermark=True, ) print(create_result) # Polling query section print("----- polling task status -----") task_id = create_result.id while True: get_result = client.content_generation.tasks.get(task_id=task_id) status = get_result.status if status == "succeeded": print("----- task succeeded -----") print(get_result) break elif status == "failed": print("----- task failed -----") print(f"Error: {get_result.error}") break else: print(f"Current status: {status}, Retrying after 30 seconds...") time.sleep(30) ``` **选项：Java** ```Java package com.ark.sample; import com.volcengine.ark.runtime.models.content_generation.*; import com.volcengine.ark.runtime.service.ArkService; import okhttp3.ConnectionPool; import okhttp3.Dispatcher; import java.util.ArrayList; import java.util.List; import java.util.concurrent.TimeUnit; public class ContentGenerationTaskExample { // Client initialization static String apiKey = System.getenv("ARK_API_KEY"); static ConnectionPool connectionPool = new ConnectionPool(5, 1, TimeUnit.SECONDS); static Dispatcher dispatcher = new Dispatcher(); static ArkService service = ArkService.builder() .baseUrl("https://ark.cn-beijing.volces.com/api/v3") // The base URL for model invocation .dispatcher(dispatcher) .connectionPool(connectionPool) .apiKey(apiKey) .build(); public static void main(String[] args) { // Model ID final String modelId = "doubao-seedance-2-0-260128"; // Text prompt final String prompt = "视频1中的拱形窗户打开，进入美术馆室内，接视频2，之后镜头进入画内，接视频3"; // Example resource URLs final String refVideo1 = "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video1.mp4"; final String refVideo2 = "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video2.mp4"; final String refVideo3 = "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video3.mp4"; // Output video parameters final boolean generateAudio = true; final String videoRatio = "16:9"; final long videoDuration = 8L; final boolean showWatermark = true; System.out.println("----- create request -----"); // Build request content List contents = new ArrayList<>(); // 1. Text prompt contents.add(ContentItem.builder() .type(ContentType.TEXT) .text(prompt) .build()); // 2. Reference video 1 contents.add(ContentItem.builder() .type(ContentType.VIDEO_URL) .videoUrl(VideoURL.builder() .url(refVideo1) .build()) .role("reference_video") .build()); // 3. Reference video 2 contents.add(ContentItem.builder() .type(ContentType.VIDEO_URL) .videoUrl(VideoURL.builder() .url(refVideo2) .build()) .role("reference_video") .build()); // 4. Reference video 3 contents.add(ContentItem.builder() .type(ContentType.VIDEO_URL) .videoUrl(VideoURL.builder() .url(refVideo3) .build()) .role("reference_video") .build()); // Create video generation task CreateContentGenerationTaskRequest createRequest = CreateContentGenerationTaskRequest.builder() .generateAudio(generateAudio) .model(modelId) .content(contents) .ratio(videoRatio) .duration(videoDuration) .watermark(showWatermark) .build(); CreateContentGenerationTaskResponse createResult = service.createContentGenerationTask(createRequest); System.out.println("Task Created: " + createResult); // Get task details and poll status String taskId = createResult.getId(); pollTaskStatus(taskId); } /** * Poll task status * @param taskId Task ID */ private static void pollTaskStatus(String taskId) { String getRequest = taskId; System.out.println("----- polling task status -----"); try { while (true) { ContentGenerationTask getResponse = service.getContentGenerationTask(getRequest); String status = getResponse.getStatus().toString(); if ("succeeded".equalsIgnoreCase(status)) { System.out.println("----- task succeeded -----"); System.out.println(getResponse); break; } else if ("failed".equalsIgnoreCase(status)) { System.out.println("----- task failed -----"); if (getResponse.getError() != null) { System.out.println("Error: " + getResponse.getError().getMessage()); } break; } else { System.out.printf("Current status: %s, Retrying in 10 seconds...%n", status); TimeUnit.SECONDS.sleep(10); } } } catch (InterruptedException ie) { Thread.currentThread().interrupt(); System.err.println("Polling interrupted"); } catch (Exception e) { System.err.println("Error occurred: " + e.getMessage()); } finally { service.shutdownExecutor(); } } } ``` **选项：Go** ```Go package main import ( "context" "fmt" "os" "time" "github.com/volcengine/ark-runtime-go/arkruntime" model "github.com/volcengine/ark-runtime-go/arkruntime/model/contentgeneration" ) func main() { // Initialize Ark client client := arkruntime.NewClientWithApiKey( os.Getenv("ARK_API_KEY"), // The base URL for model invocation arkruntime.WithBaseUrl("https://ark.cn-beijing.volces.com/api/v3"), ) ctx := context.Background() // Model ID modelID := "doubao-seedance-2-0-260128" // Text prompt prompt := "视频1中的拱形窗户打开，进入美术馆室内，接视频2，之后镜头进入画内，接视频3" // Example resource URLs refVideo1 := "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video1.mp4" refVideo2 := "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video2.mp4" refVideo3 := "https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/r2v_extend_video3.mp4" // Output video parameters generateAudio := true videoRatio := "16:9" videoDuration := int64(8) showWatermark := true // 1. Create video generation task fmt.Println("----- create request -----") createReq := &model.CreateContentGenerationTaskRequest{ Model: modelID, GenerateAudio: model.NewOptBool(generateAudio), Ratio: model.NewOptString(videoRatio), Duration: model.NewOptInt64(videoDuration), Watermark: model.NewOptBool(showWatermark), Content: []model.ContentItem{ { Type: model.ContentTypeText, Text: model.NewOptString(prompt), }, { Type: model.ContentTypeVideoURL, VideoURL: model.NewOptVideoURL(model.VideoURL{ URL: refVideo1, }), Role: model.NewOptString("reference_video"), }, { Type: model.ContentTypeVideoURL, VideoURL: model.NewOptVideoURL(model.VideoURL{ URL: refVideo2, }), Role: model.NewOptString("reference_video"), }, { Type: model.ContentTypeVideoURL, VideoURL: model.NewOptVideoURL(model.VideoURL{ URL: refVideo3, }), Role: model.NewOptString("reference_video"), }, }, } createResp, err := client.CreateContentGenerationTask(ctx, createReq) if err != nil { fmt.Printf("create content generation error: %v\n", err) return } taskID := createResp.ID fmt.Printf("Task Created with ID: %s\n", taskID) // 2. Poll task status pollTaskStatus(ctx, client, taskID) } // poll task status func pollTaskStatus(ctx context.Context, client *arkruntime.Client, taskID string) { fmt.Println("----- polling task status -----") for { getReq := taskID getResp, err := client.GetContentGenerationTask(ctx, getReq) if err != nil { fmt.Printf("get content generation task error: %v\n", err) return } status := getResp.Status if status == "succeeded" { fmt.Println("----- task succeeded -----") fmt.Printf("Task ID: %s \n", getResp.ID) fmt.Printf("Model: %s \n", getResp.Model) fmt.Printf("Video URL: %s \n", getResp.Content.Or(model.TaskContent{}).VideoURL.Or("")) fmt.Printf("Completion Tokens: %d \n", getResp.Usage.Or(model.TaskUsage{}).CompletionTokens) fmt.Printf("Created At: %d, Updated At: %d\n", getResp.CreatedAt.Or(0), getResp.UpdatedAt.Or(0)) return } else if status == "failed" { fmt.Println("----- task failed -----") if getResp.Error.IsSet() { fmt.Printf("Error Code: %s, Message: %s\n", getResp.Error.Value.Code, getResp.Error.Value.Message) } return } else { fmt.Printf("Current status: %s, Retrying in 10 seconds... \n", status) time.Sleep(10 * time.Second) } } } ``` > **说明** > > * 向前或向后延长 1 段视频，生成的视频一般只包含原视频的尾部画面。但您也可以通过提示词灵活控制，使其包含原视频内容。 例如：向前延长视频1，[延长内容描述...]，**最后接视频1**。 > * 传入 2~3 段视频，补全中间过渡部分，生成的视频会包含原视频内容和新生成的视频内容。 ## 输出 4k 视频 > 仅 Seedance 2.0 支持 Seedance 2.0 支持输出 4k 视频，并采用 10bit 位深编码，能够完整保留丰富的色彩层次与平滑的渐变过渡，满足专业影视制作与 HDR 视频内容的要求。 > **注意** > > 4k 视频采用 H.265/HEVC 编码格式输出，部分播放器或浏览器可能无法直接播放，详见 [10bit 位深与 H.265/HEVC 编码视频播放兼容性说明](/docs/seedance-2-0#4k_player)。 |效果预览1 |效果预览2 | |---|---| |[视频](https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/seedance-4k-makeup-music.mov) |[视频](https://arkdocs.tos-cn-beijing.volces.com/videos/video-generation/seedance-4k-motorcycle-music.mov) | > 注：效果展示视频由 Seedance 2.0 生成的多个分镜拼接而成，非下述示例代码直接生成。 **选项：Python** ```Python import os import time # Install SDK: python -m pip install --upgrade arkruntime from arkruntime import Ark client = Ark( # The base URL for model invocation base_url='https://ark.cn-beijing.volces.com/api/v3', # Get API Key：https://ark.volcengine.com/region:cn-beijing/apikey api_key=os.environ.get("ARK_API_KEY"), ) if __name__ == "__main__": print("----- create request -----") create_result = client.content_generation.tasks.create( model="doubao-seedance-2-0-260128", content=[ { "type": "text", "text": "生成一段15秒的越野摩托竞技广告感短片。参考图片作为中段飞跃高潮的参考。镜头逻辑依次为：1）中景跟拍，车手从远处沿土坡高速逼近跳台；2）超近低机位后轮飞砂特写，轮胎抓地甩出大量泥土和砂石；3）中近景展示骑手控车、手部发力、悬挂压缩与机械震动；4）侧向英雄中景拍车手冲坡腾空飞跃，画面状态接近图一，泥土在逆光中大面积飞散；5）腾空近景帅气细节，突出头盔护目镜、手部控把、轮胎悬空或车身侧面局部；6）中景跟拍落地，悬挂压缩回弹，随后继续沿土坡赛道高速冲刺收尾。全片同一名骑手、同一辆车、同一条赛道，镜头景别和角度区分清楚，不重复，动作连贯,画面有真实越野跟拍抖动感、速度感、扬土感和夕阳逆光竞技氛围。", }, { "type": "image_url", "image_url": { "url": "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/i2v_4k.png" }, "role": "reference_image", }, ], generate_audio=True, resolution="4k", ratio="adaptive", duration=15, watermark=True, ) print(create_result) # Polling query section print("----- polling task status -----") task_id = create_result.id while True: get_result = client.content_generation.tasks.get(task_id=task_id) status = get_result.status if status == "succeeded": print("----- task succeeded -----") print(get_result) break elif status == "failed": print("----- task failed -----") print(f"Error: {get_result.error}") break else: print(f"Current status: {status}, Retrying after 30 seconds...") time.sleep(30) ``` **选项：Java** ```Java package com.ark.sample; import com.volcengine.ark.runtime.models.content_generation.*; import com.volcengine.ark.runtime.service.ArkService; import okhttp3.ConnectionPool; import okhttp3.Dispatcher; import java.util.ArrayList; import java.util.List; import java.util.concurrent.TimeUnit; public class ContentGenerationTaskExample { // Client initialization static String apiKey = System.getenv("ARK_API_KEY"); static ConnectionPool connectionPool = new ConnectionPool(5, 1, TimeUnit.SECONDS); static Dispatcher dispatcher = new Dispatcher(); static ArkService service = ArkService.builder() .baseUrl("https://ark.cn-beijing.volces.com/api/v3") // The base URL for model invocation .dispatcher(dispatcher) .connectionPool(connectionPool) .apiKey(apiKey) .build(); public static void main(String[] args) { // Model ID final String modelId = "doubao-seedance-2-0-260128"; // Text prompt final String prompt = "生成一段15秒的越野摩托竞技广告感短片。参考图片作为中段飞跃高潮的参考。" + "镜头逻辑依次为：1）中景跟拍，车手从远处沿土坡高速逼近跳台；" + "2）超近低机位后轮飞砂特写，轮胎抓地甩出大量泥土和砂石；" + "3）中近景展示骑手控车、手部发力、悬挂压缩与机械震动；" + "4）侧向英雄中景拍车手冲坡腾空飞跃，画面状态接近图一，泥土在逆光中大面积飞散；" + "5）腾空近景帅气细节，突出头盔护目镜、手部控把、轮胎悬空或车身侧面局部；" + "6）中景跟拍落地，悬挂压缩回弹，随后继续沿土坡赛道高速冲刺收尾。" + "全片同一名骑手、同一辆车、同一条赛道，镜头景别和角度区分清楚，不重复，动作连贯,画面有真实越野跟拍抖动感、速度感、扬土感和夕阳逆光竞技氛围。"; // Example resource URLs final String refImage = "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/i2v_4k.png"; // Output video parameters final boolean generateAudio = true; final String videoResolution = "4k"; final String videoRatio = "adaptive"; final long videoDuration = 15L; final boolean showWatermark = true; System.out.println("----- create request -----"); // Build request content List contents = new ArrayList<>(); // 1. Text prompt contents.add(ContentItem.builder() .type(ContentType.TEXT) .text(prompt) .build()); // 2. Reference image contents.add(ContentItem.builder() .type(ContentType.IMAGE_URL) .imageUrl(ImageURL.builder() .url(refImage) .build()) .role("reference_image") .build()); // Create video generation task CreateContentGenerationTaskRequest createRequest = CreateContentGenerationTaskRequest.builder() .generateAudio(generateAudio) .model(modelId) .content(contents) .resolution(videoResolution) .ratio(videoRatio) .duration(videoDuration) .watermark(showWatermark) .build(); CreateContentGenerationTaskResponse createResult = service.createContentGenerationTask(createRequest); System.out.println("Task Created: " + createResult); // Get task details and poll status String taskId = createResult.getId(); pollTaskStatus(taskId); } /** * Poll task status * @param taskId Task ID */ private static void pollTaskStatus(String taskId) { String getRequest = taskId; System.out.println("----- polling task status -----"); try { while (true) { ContentGenerationTask getResponse = service.getContentGenerationTask(getRequest); String status = getResponse.getStatus().toString(); if ("succeeded".equalsIgnoreCase(status)) { System.out.println("----- task succeeded -----"); System.out.println(getResponse); break; } else if ("failed".equalsIgnoreCase(status)) { System.out.println("----- task failed -----"); if (getResponse.getError() != null) { System.out.println("Error: " + getResponse.getError().getMessage()); } break; } else { System.out.printf("Current status: %s, Retrying in 10 seconds...%n", status); TimeUnit.SECONDS.sleep(10); } } } catch (InterruptedException ie) { Thread.currentThread().interrupt(); System.err.println("Polling interrupted"); } catch (Exception e) { System.err.println("Error occurred: " + e.getMessage()); } finally { service.shutdownExecutor(); } } } ``` **选项：Go** ```Go package main import ( "context" "fmt" "os" "time" "github.com/volcengine/ark-runtime-go/arkruntime" model "github.com/volcengine/ark-runtime-go/arkruntime/model/contentgeneration" ) func main() { // Initialize Ark client client := arkruntime.NewClientWithApiKey( os.Getenv("ARK_API_KEY"), // The base URL for model invocation arkruntime.WithBaseUrl("https://ark.cn-beijing.volces.com/api/v3"), ) ctx := context.Background() // Model ID modelID := "doubao-seedance-2-0-260128" // Text prompt prompt := "生成一段15秒的越野摩托竞技广告感短片。参考图片作为中段飞跃高潮的参考。" + "镜头逻辑依次为：1）中景跟拍，车手从远处沿土坡高速逼近跳台；" + "2）超近低机位后轮飞砂特写，轮胎抓地甩出大量泥土和砂石；" + "3）中近景展示骑手控车、手部发力、悬挂压缩与机械震动；" + "4）侧向英雄中景拍车手冲坡腾空飞跃，画面状态接近图一，泥土在逆光中大面积飞散；" + "5）腾空近景帅气细节，突出头盔护目镜、手部控把、轮胎悬空或车身侧面局部；" + "6）中景跟拍落地，悬挂压缩回弹，随后继续沿土坡赛道高速冲刺收尾。" + "全片同一名骑手、同一辆车、同一条赛道，镜头景别和角度区分清楚，不重复，动作连贯,画面有真实越野跟拍抖动感、速度感、扬土感和夕阳逆光竞技氛围。" // Example resource URLs refImage := "https://arkdocs.tos-cn-beijing.volces.com/images/video-generation/i2v_4k.png" // Output video parameters generateAudio := true videoResolution := "4k" videoRatio := "adaptive" videoDuration := int64(15) showWatermark := true // 1. Create video generation task fmt.Println("----- create request -----") createReq := &model.CreateContentGenerationTaskRequest{ Model: modelID, GenerateAudio: model.NewOptBool(generateAudio), Resolution: model.NewOptString(videoResolution), Ratio: model.NewOptString(videoRatio), Duration: model.NewOptInt64(videoDuration), Watermark: model.NewOptBool(showWatermark), Content: []model.ContentItem{ { Type: model.ContentTypeText, Text: model.NewOptString(prompt), }, { Type: model.ContentTypeImageURL, ImageURL: model.NewOptImageURL(model.ImageURL{ URL: refImage, }), Role: model.NewOptString("reference_image"), }, }, } createResp, err := client.CreateContentGenerationTask(ctx, createReq) if err != nil { fmt.Printf("create content generation error: %v\n", err) return } taskID := createResp.ID fmt.Printf("Task Created with ID: %s\n", taskID) // 2. Poll task status pollTaskStatus(ctx, client, taskID) } // poll task status func pollTaskStatus(ctx context.Context, client *arkruntime.Client, taskID string) { fmt.Println("----- polling task status -----") for { getReq := taskID getResp, err := client.GetContentGenerationTask(ctx, getReq) if err != nil { fmt.Printf("get content generation task error: %v\n", err) return } status := getResp.Status if status == "succeeded" { fmt.Println("----- task succeeded -----") fmt.Printf("Task ID: %s \n", getResp.ID) fmt.Printf("Model: %s \n", getResp.Model) fmt.Printf("Video URL: %s \n", getResp.Content.Or(model.TaskContent{}).VideoURL.Or("")) fmt.Printf("Completion Tokens: %d \n", getResp.Usage.Or(model.TaskUsage{}).CompletionTokens) fmt.Printf("Created At: %d, Updated At: %d\n", getResp.CreatedAt.Or(0), getResp.UpdatedAt.Or(0)) return } else if status == "failed" { fmt.Println("----- task failed -----") if getResp.Error.IsSet() { fmt.Printf("Error Code: %s, Message: %s\n", getResp.Error.Value.Code, getResp.Error.Value.Message) } return } else { fmt.Printf("Current status: %s, Retrying in 10 seconds... \n", status) time.Sleep(10 * time.Second) } } } ``` ## 使用联网搜索 > 联网搜索能力仅适用于纯文本输入 通过配置 tools.**type** 参数为`web_search`即可使用联网搜索工具。 * 开启联网搜索后，模型会根据用户的提示词自主判断是否搜索互联网内容（如商品、天气等）。可提升生成视频的时效性，但也会增加一定的时延。 * 实际搜索次数可通过 [查询视频生成任务 API](/docs/get-video-generation-task-api) 返回的 usage.tool_usage.**web_search** 字段获取，如果为 0 表示未搜索。 |输入：文本 |输出 | |---|---| |微距镜头对准叶片上翠绿的玻璃蛙。焦点逐渐从它光滑的皮肤，转移到它完全透明的腹部，一颗鲜红的心脏正在有力地、规律地收缩扩张。 |[视频](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/afad79fc76a34d1fbe7b2c809d1e19f1) | \ | | | \ |:::tip | | \ |联网搜索玻璃蛙的容貌特征。 | | \ |::: | | **选项：Python** ```Python import os import time # Install SDK: python -m pip install --upgrade arkruntime from arkruntime import Ark # Make sure that you have stored the API Key in the environment variable ARK_API_KEY # Initialize the Ark client to read your API Key from an environment variable client = Ark( # This is the default path. You can configure it based on the service location base_url="https://ark.cn-beijing.volces.com/api/v3", # Get API Key：https://ark.volcengine.com/region:cn-beijing/apikey api_key=os.environ.get("ARK_API_KEY"), ) if __name__ == "__main__": print("----- create request -----") create_result = client.content_generation.tasks.create( model="doubao-seedance-2-0-260128", # Replace with Model ID content=[ { # text prompt "type": "text", "text": "微距镜头对准叶片上翠绿的玻璃蛙。焦点逐渐从它光滑的皮肤，转移到它完全透明的腹部，一颗鲜红的心脏正在有力地、规律地收缩扩张。" } ], generate_audio=True, ratio="16:9", duration=11, watermark=True, tools=[{"type": "web_search"}], ) print(create_result) # Polling query section print("----- polling task status -----") task_id = create_result.id while True: get_result = client.content_generation.tasks.get(task_id=task_id) status = get_result.status if status == "succeeded": print("----- task succeeded -----") print(get_result) break elif status == "failed": print("----- task failed -----") print(f"Error: {get_result.error}") break else: print(f"Current status: {status}, Retrying after 10 seconds...") time.sleep(10) ``` **选项：Java** ```Java package com.ark.sample; import com.volcengine.ark.runtime.models.content_generation.*; import com.volcengine.ark.runtime.service.ArkService; import okhttp3.ConnectionPool; import okhttp3.Dispatcher; import java.util.ArrayList; import java.util.List; import java.util.concurrent.TimeUnit; import java.util.Collections; public class ContentGenerationTaskExample { // Make sure that you have stored the API Key in the environment variable ARK_API_KEY // Initialize the Ark client to read your API Key from an environment variable static String apiKey = System.getenv("ARK_API_KEY"); static ConnectionPool connectionPool = new ConnectionPool(5, 1, TimeUnit.SECONDS); static Dispatcher dispatcher = new Dispatcher(); static ArkService service = ArkService.builder() .baseUrl("https://ark.cn-beijing.volces.com/api/v3") // The base URL for model invocation .dispatcher(dispatcher) .connectionPool(connectionPool) .apiKey(apiKey) .build(); public static void main(String[] args) { String model = "doubao-seedance-2-0-260128"; // Replace with Model ID String prompt = "微距镜头对准叶片上翠绿的玻璃蛙。焦点逐渐从它光滑的皮肤，转移到它完全透明的腹部，一颗鲜红的心脏正在有力地、规律地收缩扩张。"; Boolean generateAudio = true; String videoRatio = "16:9"; Long videoDuration = 11L; Boolean showWatermark = true; // Create ContentGenerationTool Tool webSearchTool = Tool.builder().type(ToolType.WEB_SEARCH).build(); System.out.println("----- create request -----"); List contents = new ArrayList<>(); // text prompt contents.add(ContentItem.builder() .type(ContentType.TEXT) .text(prompt) .build()); // Create a video generation task CreateContentGenerationTaskRequest createRequest = CreateContentGenerationTaskRequest.builder() .model(model) .content(contents) .generateAudio(generateAudio) .ratio(videoRatio) .duration(videoDuration) .watermark(showWatermark) .tools(Collections.singletonList(webSearchTool)) .build(); CreateContentGenerationTaskResponse createResult = service.createContentGenerationTask(createRequest); System.out.println(createResult); // Get the details of the task String taskId = createResult.getId(); String getRequest = taskId; // Polling query section System.out.println("----- polling task status -----"); while (true) { try { ContentGenerationTask getResponse = service.getContentGenerationTask(getRequest); String status = getResponse.getStatus().toString(); if ("succeeded".equalsIgnoreCase(status)) { System.out.println("----- task succeeded -----"); System.out.println(getResponse); break; } else if ("failed".equalsIgnoreCase(status)) { System.out.println("----- task failed -----"); System.out.println("Error: " + getResponse.getStatus()); break; } else { System.out.printf("Current status: %s, Retrying in 10 seconds...", status); TimeUnit.SECONDS.sleep(10); } } catch (InterruptedException ie) { Thread.currentThread().interrupt(); System.err.println("Polling interrupted"); break; } } } } ``` **选项：Go** ```Go package main import ( "context" "fmt" "os" "time" "github.com/volcengine/ark-runtime-go/arkruntime" model "github.com/volcengine/ark-runtime-go/arkruntime/model/contentgeneration" ) func main() { // Make sure that you have stored the API Key in the environment variable ARK_API_KEY // Initialize the Ark client to read your API Key from an environment variable client := arkruntime.NewClientWithApiKey( // Get your API Key from the environment variable. This is the default mode and you can modify it as required os.Getenv("ARK_API_KEY"), // The base URL for model invocation arkruntime.WithBaseUrl("https://ark.cn-beijing.volces.com/api/v3"), ) ctx := context.Background() // Model ID modelID := "doubao-seedance-2-0-260128" // Text prompt prompt := "微距镜头对准叶片上翠绿的玻璃蛙。焦点逐渐从它光滑的皮肤，转移到它完全透明的腹部，一颗鲜红的心脏正在有力地、规律地收缩扩张。" // Output video parameters generateAudio := true videoRatio := "16:9" videoDuration := int64(11) showWatermark := true // Create ContentGenerationTool tools := []model.Tool{ {Type: model.ToolTypeWebSearch}, } // Generate a task fmt.Println("----- create request -----") createReq := &model.CreateContentGenerationTaskRequest{ Model: modelID, GenerateAudio: model.NewOptBool(generateAudio), Ratio: model.NewOptString(videoRatio), Duration: model.NewOptInt64(videoDuration), Watermark: model.NewOptBool(showWatermark), Tools: tools, Content: []model.ContentItem{ { // Combination of text prompt and parameters Type: model.ContentTypeText, Text: model.NewOptString(prompt), }, }, } createResp, err := client.CreateContentGenerationTask(ctx, createReq) if err != nil { fmt.Printf("create content generation error: %v\n", err) return } taskID := createResp.ID fmt.Printf("Task Created with ID: %s\n", taskID) // 2. Poll task status pollTaskStatus(ctx, client, taskID) } // poll task status func pollTaskStatus(ctx context.Context, client *arkruntime.Client, taskID string) { fmt.Println("----- polling task status -----") for { getReq := taskID getResp, err := client.GetContentGenerationTask(ctx, getReq) if err != nil { fmt.Printf("get content generation task error: %v\n", err) return } status := getResp.Status if status == "succeeded" { fmt.Println("----- task succeeded -----") fmt.Printf("Task ID: %s \n", getResp.ID) fmt.Printf("Model: %s \n", getResp.Model) fmt.Printf("Video URL: %s \n", getResp.Content.Or(model.TaskContent{}).VideoURL.Or("")) fmt.Printf("Completion Tokens: %d \n", getResp.Usage.Or(model.TaskUsage{}).CompletionTokens) fmt.Printf("Created At: %d, Updated At: %d\n", getResp.CreatedAt.Or(0), getResp.UpdatedAt.Or(0)) return } else if status == "failed" { fmt.Println("----- task failed -----") if getResp.Error.IsSet() { fmt.Printf("Error Code: %s, Message: %s\n", getResp.Error.Value.Code, getResp.Error.Value.Message) } return } else { fmt.Printf("Current status: %s, Retrying in 10 seconds... \n", status) time.Sleep(10 * time.Second) } } } ``` ## 更多能力 Seedance 2.0 系列模型也支持文生视频、首帧图生视频、首尾帧图生视频、自定义视频输出规格（包括：分辨率、宽高比、视频时长、视频中是否包含水印）等通用基础能力，详情请参见 [视频生成教程](/docs/video-generation-tutorial)。 # 便利创作 Seedance 2.0 系列模型不支持直接上传含有真人人脸的参考图/视频。为便利创作者使用肖像，平台推出了以下解决方案，详细教程请参见 [Doubao Seedance 便利创作含肖像视频](/docs/seedance-portrait-asset-guide)： |方案 |介绍 | |---|---| |[信任模型产物作为输入素材](/docs/seedance-portrait-asset-guide#trust-model-output) | 本账号下部分模型生成的含人脸原始产物可作为输入素材，再次调用 Seedance 2.0 系列模型进行二次创作，不会触发输入审核拦截。 | | [使用预置虚拟人像](/docs/seedance-portrait-asset-guide#preset-avatar) |平台预置虚拟人像库，为创作者提供免费、合规、丰富多样的肖像素材。适用于需真人风格人脸但无需指定具体人物，追求零合规风险、快速创作的场景。 | |[使用已授权真人素材](/docs/seedance-portrait-asset-guide#authorized-real-person) |支持使用已获得授权的真人肖像素材进行视频生成。 | # 提示词技巧 ## 提示词 Skill 平台提供 **Seedance 2.0 提示词优化技能**，方便您对提示词进行调优。 * **配置方式**：推荐在本地项目中运行以下命令，通过方舟文档 Skills catalog 安装 `sd2-pe`： ```bash npx --yes skills@latest add "https://arkdocs.tos-cn-beijing.volces.com/skills/" \ --skill sd2-pe \ --yes ``` 如需手动安装，可以下载以下 `SKILL.md` 文件。以 OpenClaw 为例，将文件完整内容复制到对话输入框中，并发送“请帮我安装这个技能”，等待工具完成安装。 [SKILL.md](https://arkdocs.tos-cn-beijing.volces.com/files/video-generation/SKILL.md) * **使用方式**：在 AI 对话框输入 `/sd2-pe + 你的提示词内容`，开始调试提示词。 ## 提示词规则 - 提示词中必须使用"**素材类型+序号**”格式引用素材，序号为请求体中该素材在同类素材中的排序。例如 「图片 n」指代`content`数组中第 n 个`type="image_url"`的参考图片（按数组顺序从1开始计数）。**注意不支持使用 Asset ID 指代素材**。 - 不同任务的提示词公式及详细规则请参见 [Doubao Seedance 2.0 系列提示词指南](/docs/seedance-2-0-prompt-guide)。 # 使用限制 参见 [使用限制](/docs/seedance-2-0#66cb028f)。 # 常见问题 ## 10bit 位深与 H.265/HEVC 编码视频播放兼容性说明 以下为各平台浏览器及播放器对 10bit 位深与 H.265/HEVC 编码视频的播放兼容性测试结论，实际效果可能因设备配置有所差异。 **推荐使用：** - **macOS 端**：浏览器推荐 Safari、Chrome，播放器推荐 VLC、mpv、QuickTime Player - **Windows 端**：浏览器推荐 Edge、Chrome，播放器推荐 VLC、mpv ### Windows 平台 **选项：浏览器** | 浏览器 | 支持情况 | |------|---------| | Chrome | 有条件支持 | | Edge | 有条件支持 | | Firefox | 有条件支持 | | 360 浏览器 | 有条件支持 | | QQ 浏览器 | 有条件支持 | | Opera | 有条件支持 | **选项：播放器** | 播放器 | 支持情况 | |------|---------| | VLC | 支持 | | 系统「电影和电视」 | 有条件支持 | | PotPlayer | 有条件支持 | | 迅雷影音 | 支持 | | QQ 影音 | 有条件支持 | | MPC-HC / MPC-BE | 有条件支持 | | mpv | 支持 | | KMPlayer | 支持 | > **有条件支持说明**：需具备较高的硬件解码能力。已知在 Intel i7 + NVIDIA RTX 4070 + Windows 11 及更高配置下可正常播放，其他配置建议以实际测试为准。 ### macOS 平台 **选项：浏览器** | 浏览器 | 支持情况 | |------|---------| | Safari | 支持 | | Chrome | 有条件支持 | | Edge | 有条件支持 | | Firefox | 有条件支持 | | Opera | 有条件支持 | **选项：播放器** | 播放器 | 支持情况 | |------|---------| | VLC | 支持 | | QuickTime Player | 支持 | | IINA | 支持 | | mpv | 支持 | | Infuse | 支持 | | Kodi | 有条件支持 | > **有条件支持说明**：需具备较高的硬件解码能力。已知 Apple M2 及以上机型可正常播放，M1 及以下机型建议以实际测试为准。 ## 视频画面存在跳变 **典型现象** **首帧图生视频**、**首尾帧图生视频**场景中，生成视频部分帧出现画面拉伸、压缩等跳变问题。 **根因分析** 输入图片与输出视频的分辨率宽高不一致，引发视频画面帧间跳变。 **解决方案** 1. 裁剪输入图片：参考 Seedance 2.0 系列模型支持的宽高像素值表格（见 [创建视频生成任务 API](/docs/create-video-generation-task-api) ratio 字段），将输入图片裁剪为目标宽高像素值。 2. 将 API 的 **ratio** 字段设置为`adaptive`。 3. 使用 Seedance 2.0 系列模型重新发起首帧/首尾帧图生视频任务。