Seedream 4 API
BytePlus 统一多模态模型:使用 Seedream 4 API 生成连贯的视觉叙事内容。锁定角色形象,融合多种风格,在数秒内生成最长 15 帧的画面序列。
A beautiful landscape with mountains and a lake, golden hour lighting, photorealistic style

Upload reference images
Click to upload or drag and drop
Supported formats: JPG, JPEG, PNG, WEBP
Maximum file size: 10MB; Maximum files: 10

History
最多保留20条0 运行中 · 0 已完成
Pricing
Seedream 4 API:打造一致视觉叙事的引擎
Seedream 4 API 专为需要跨图像一致性的工作流设计——分镜、角色驱动的营销活动和产品可视化。启用序列生成时,每次请求最多可生成 15 张图像,通过多参考输入约束身份/风格,并支持高分辨率输出。实际效果因提示词和配置而异。

使用 Seedream 4 API 可以构建什么?
为何开发者选择 Seedream 4 API
专为要求高一致性、高速度和超大上下文记忆的生产级流水线而设计。
统一架构
在一次模型调用中同时完成文生图与图生图编辑,降低延迟和整体成本。
深度参考融合
上传最多 10 张参考图,强约束风格与角色身份,大幅降低创作漂移。
生产级速度
Seedream 4 针对低延迟生成进行了优化。实际延迟因输出尺寸、批量设置和路由而异——请在控制台中查看。
如何接入 Seedream 4 API
用于生成一致图像序列的精简工作流。
步骤 1 — 定义参考图
向 Seedream 4 API 发送最多 10 个图片 URL,用于确定“锚点”角色或风格。
步骤 2 — 设置序列模式
启用 `sequential_image_generation='auto'`,并在提示词中写出叙事关键节点,以触发批量生成。
步骤 3 — 获取并微调
接收最多 15 张保持高度一致性的图像 JSON 返回结果,如有需要可调用编辑端点对局部细节进行修正。
Seedream 4 API 能力
可扩展图像生成的技术规格
15 帧批量生成
在一次 API 调用中生成整组场景画面。
多参考融合
融合最多 10 个输入,获得更精细的画面控制。
身份锁定
在不同角度和镜头下保持面部特征一致。
原生 4K
通过 EvoLink 输出超高分辨率图像。
智能擦除与修补
使用自然语言编辑指定区域,无需复杂蒙版操作。
低延迟
针对响应式生成进行优化。实际延迟取决于尺寸、批量数和路由——请在控制台中查看。
Seedream 4 API 对比竞品
为你的自动化流水线选择合适的模型
| 模型 | 类别 | 核心能力 | 输出尺寸 | 计费方式 | 最适合 |
|---|---|---|---|---|---|
| Seedream 4 (BytePlus) | 图像生成 + 编辑 | 多图批量生成(最多 15 张)、多参考融合、一致性序列 | 2K / 4K 可选(取决于配置) | 按图计费(用量制) | 分镜、角色一致性,以及需要跨镜头稳定视觉的多图工作流 |
| Nano Banana 2 (Gemini Flash Image) | 快速单图生成 | 低延迟通用生成 | 1024px 基础(取决于配置) | 用量制 | 速度优先于多帧一致性的单图生成场景 |
| Wan 2.5 (Text-to-Image) | 图像生成 | 强提示词遵从的创意生成(取决于变体) | 1024–1536px(取决于配置) | 用量制 | 不需要多帧序列一致性的通用创意生成 |
Seedream 4 API 常见问题
Everything you need to know about the product and billing.
API Reference
Select endpoint
Authentication
All APIs require Bearer Token authentication.
Authorization:
Bearer YOUR_API_KEY/v1/images/generationsGenerate Image
Create an image generation task using text prompts or reference images. Supports text-to-image, image-to-image, and image editing modes.
Asynchronous processing mode, use the returned task ID to .
Generated image links are valid for 24 hours, please save them promptly.
Request Parameters
modelstringRequiredDefault: doubao-seedream-4.0Image generation model name.
doubao-seedream-4.0promptstringRequiredPrompt describing the image to be generated or how to edit the input image.
Notes
- Limited to 2000 tokens
A cat playing in the grasssizestringOptionalSize of generated images. Supports simplified format or pixel format.
| Value | Description |
|---|---|
| 1K | Standard quality |
| 2K | High quality |
| 4K | Ultra quality |
Notes
- Pixel format: 1280x720 to 4096x4096
- Aspect ratio range: [1/16, 16]
2048x2048nintegerOptionalDefault: 1Number of images to generate.
| Value | Description |
|---|---|
| 1-15 | Batch generation supported |
1image_urlsarrayOptionalReference image URL list for image-to-image and image editing functions.
Notes
- Max 10 images per request
- Max size: 10MB per image
- Formats: .jpeg, .jpg, .png
- Aspect ratio range: [1/3, 3]
- Total pixels: ≤6000×6000
- URLs must be directly accessible by the server
https://example.com/image1.pngprompt_priorityenum<string>OptionalDefault: standardPrompt optimization strategy for setting the prompt optimization mode.
| Value | Description |
|---|---|
| standard | Standard mode, higher quality, slower |
| fast | Fast mode, faster speed, normal quality |
standardcallback_urlstringOptionalHTTPS callback address after task completion.
Notes
- Triggered on completion, failure, or cancellation
- Sent after billing confirmation
- HTTPS only, no internal IPs
- Max length: 2048 chars
- Timeout: 10s, Max 3 retries
https://your-domain.com/webhooks/image-task-completed

