Seedream v4.5 Sequential API by ByteDance
ByteDance latest image generation model with batch generation support. Generate up to 15 images in a single request.
代码示例
import requests
import time
# Step 1: Start image generation
generate_url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
"Content-Type": "application/json",
"Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
"model": "bytedance/seedream-v4.5/sequential",
"prompt": "A beautiful landscape with mountains and lake",
"width": 512,
"height": 512,
"steps": 20,
"guidance_scale": 7.5,
}
generate_response = requests.post(generate_url, headers=headers, json=data)
generate_result = generate_response.json()
prediction_id = generate_result["data"]["id"]
# Step 2: Poll for result
poll_url = f"https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}"
def check_status():
while True:
response = requests.get(poll_url, headers={"Authorization": "Bearer $ATLASCLOUD_API_KEY"})
result = response.json()
if result["data"]["status"] == "completed":
print("Generated image:", result["data"]["outputs"][0])
return result["data"]["outputs"][0]
elif result["data"]["status"] == "failed":
raise Exception(result["data"]["error"] or "Generation failed")
else:
# Still processing, wait 2 seconds
time.sleep(2)
image_url = check_status()安装
安装所需的依赖包。
pip install requests认证
所有 API 请求需要通过 API Key 进行认证。您可以在 Atlas Cloud 控制台获取 API Key。
export ATLASCLOUD_API_KEY="your-api-key-here"HTTP 请求头
import os
API_KEY = os.environ.get("ATLASCLOUD_API_KEY")
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {API_KEY}"
}切勿在客户端代码或公开仓库中暴露您的 API Key。请使用环境变量或后端代理。
提交请求
import requests
url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
"Content-Type": "application/json",
"Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
"model": "your-model",
"prompt": "A beautiful landscape"
}
response = requests.post(url, headers=headers, json=data)
print(response.json())提交请求
提交一个异步生成请求。API 返回一个 prediction ID，您可以用它来检查状态和获取结果。
/api/v1/model/generateImage请求体
import requests
url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
"Content-Type": "application/json",
"Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
"model": "bytedance/seedream-v4.5/sequential",
"input": {
"prompt": "A beautiful landscape with mountains and lake"
}
}
response = requests.post(url, headers=headers, json=data)
result = response.json()
print(f"Prediction ID: {result['id']}")
print(f"Status: {result['status']}")响应
{
"id": "pred_abc123",
"status": "processing",
"model": "model-name",
"created_at": "2025-01-01T00:00:00Z"
}检查状态
轮询 prediction 端点以检查请求的当前状态。
/api/v1/model/prediction/{prediction_id}轮询示例
import requests
import time
prediction_id = "pred_abc123"
url = f"https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}"
headers = { "Authorization": "Bearer $ATLASCLOUD_API_KEY" }
while True:
response = requests.get(url, headers=headers)
result = response.json()
status = result["data"]["status"]
print(f"Status: {status}")
if status in ["completed", "succeeded"]:
output_url = result["data"]["outputs"][0]
print(f"Output URL: {output_url}")
break
elif status == "failed":
print(f"Error: {result['data'].get('error', 'Unknown')}")
break
time.sleep(3)状态值
processing请求仍在处理中。completed生成完成，输出可用。succeeded生成成功，输出可用。failed生成失败，请检查 error 字段。完成响应
{
"data": {
"id": "pred_abc123",
"status": "completed",
"outputs": [
"https://storage.atlascloud.ai/outputs/result.png"
],
"metrics": {
"predict_time": 8.3
},
"created_at": "2025-01-01T00:00:00Z",
"completed_at": "2025-01-01T00:00:10Z"
}
}上传文件
将文件上传到 Atlas Cloud 存储，获取可在 API 请求中使用的 URL。使用 multipart/form-data 上传。
/api/v1/model/uploadMedia上传示例
import requests
url = "https://api.atlascloud.ai/api/v1/model/uploadMedia"
headers = { "Authorization": "Bearer $ATLASCLOUD_API_KEY" }
with open("image.png", "rb") as f:
files = {"file": ("image.png", f, "image/png")}
response = requests.post(url, headers=headers, files=files)
result = response.json()
download_url = result["data"]["download_url"]
print(f"File URL: {download_url}")响应
{
"data": {
"download_url": "https://storage.atlascloud.ai/uploads/abc123/image.png",
"file_name": "image.png",
"content_type": "image/png",
"size": 1024000
}
}Input Schema
以下参数在请求体中被接受。
暂无可用参数。
请求体示例
{
"model": "bytedance/seedream-v4.5/sequential"
}Output Schema
API 返回包含生成输出 URL 的 prediction 响应。
响应示例
{
"id": "pred_abc123",
"status": "completed",
"model": "model-name",
"outputs": [
"https://storage.atlascloud.ai/outputs/result.png"
],
"metrics": {
"predict_time": 8.3
},
"created_at": "2025-01-01T00:00:00Z",
"completed_at": "2025-01-01T00:00:10Z"
}Atlas Cloud Skills
Atlas Cloud Skills 将 300+ AI 模型直接集成到您的 AI 编程助手中。一条命令安装，即可用自然语言生成图像、视频和与 LLM 对话。
支持的客户端
安装
npx skills add AtlasCloudAI/atlas-cloud-skills设置 API Key
从 Atlas Cloud 控制台获取 API Key，并将其设置为环境变量。
export ATLASCLOUD_API_KEY="your-api-key-here"功能
安装后，您可以在 AI 助手中使用自然语言访问所有 Atlas Cloud 模型。
MCP Server
Atlas Cloud MCP Server 通过 Model Context Protocol 将您的 IDE 与 300+ AI 模型连接。支持任何兼容 MCP 的客户端。
支持的客户端
安装
npx -y atlascloud-mcp配置
将以下配置添加到您的 IDE 的 MCP 设置文件中。
{
"mcpServers": {
"atlascloud": {
"command": "npx",
"args": [
"-y",
"atlascloud-mcp"
],
"env": {
"ATLASCLOUD_API_KEY": "your-api-key-here"
}
}
}
}可用工具
API Schema
Schema 不可用Seedream下一代视觉创作
ByteDance 最先进的图像生成模型,具备卓越美学、更高一致性和更智能的指令遵循能力。
主要更新
体验下一代 AI 驱动的视觉创作
卓越美学
生成电影级视觉效果,精细的光影和渲染,达到专业级输出。
更高一致性
在多张图像中保持稳定的主体、清晰的细节和连贯的场景。
更智能的指令遵循
准确响应复杂提示词,具备精确的视觉控制和交互式编辑能力。
更强的空间理解
生成逼真的比例、物体摆放和场景布局,准确度更高。
更丰富的世界知识
创建基于知识的视觉内容,具备准确的科学和技术推理能力。
更深度的行业应用
支持电商、影视、广告、游戏等领域的专业工作流程。
行业应用
电商
产品摄影与营销
影视
概念设计与分镜
广告
营销视觉与创意
游戏
角色与场景设计
教育
教学插图
室内设计
空间可视化
建筑
建筑渲染
时尚
虚拟试穿与造型
相比 4.0 的改进
查看 Seedream 4.5 如何超越前一版本
面部质量
面部比例较小时有显著改进
文字渲染
增强的小字符渲染能力
身份保留
更强的身份保持能力
准备开始创作?
体验 Seedream 4.5 的强大功能,改变您的创意工作流程。
Seedream 4.5 : A professional, high-fidelity multimodal image generation model by ByteDance Seed
Model Card Overview
| Field | Description |
|---|---|
| Model Name | Seedream 4.5 |
| Developed By | ByteDance Seed |
| Release Date | December 2025 |
| Model Type | Multimodal Image Generation |
| Related Links | Official Website,Technical Paper (arXiv), GitHub Repository |
Introduction
Seedream 4.5 is a state-of-the-art, multimodal generative model engineered for scalability, efficiency, and professional-grade output. As an advanced version of Seedream 4.0, it is built upon a unified framework that seamlessly integrates text-to-image synthesis, sophisticated image editing, and complex multi-image composition. The model's primary design goal is to deliver professional visual creatives with exceptional consistency and fidelity. This is achieved through a significant scaling of the model architecture and training data, which enhances its ability to preserve reference details, render dense text and typography accurately, and understand nuanced user instructions.
Key Features & Innovations
- Unified Multimodal Framework: Integrates text-to-image (T2I), single-image editing, and multi-image composition into a single, cohesive model, allowing for diverse and flexible creative workflows.
- High-Fidelity & High-Resolution Generation: Capable of generating native high-resolution images (up to 4K), capturing fine details, realistic textures, and accurate lighting for professional use cases.
- Advanced Image Editing: Excels at preserving the core structure, lighting, and color tone of reference images while applying precise edits based on natural language instructions.
- Enhanced Multi-Image Composition: Accurately identifies and blends main subjects from multiple reference images, enabling complex creative compositions and style fusions.
- Superior Typography and Text Rendering: Features significantly improved capabilities for rendering clear, legible, and contextually integrated text within images.
- Efficient and Scalable Architecture: Built on a highly efficient Diffusion Transformer (DiT) and a powerful Variational Autoencoder (VAE), enabling fast inference and effective scalability.
- Optimized for Professional Use: Demonstrates strong performance in generating structured, knowledge-based content such as design materials, posters, and product visualizations, bridging the gap between creative generation and practical industry applications.
Model Architecture & Technical Details
Seedream 4.5's architecture is an extension of the foundation laid by Seedream 4.0. The core of the model is a highly efficient and scalable Diffusion Transformer (DiT), which significantly increases model capacity while reducing computational requirements for training and inference. This is paired with a powerful Variational Autoencoder (VAE) with a high compression ratio, which minimizes the number of image tokens processed in the latent space, further boosting efficiency.
Training and Data: The model was pre-trained on billions of text-image pairs, covering a vast range of taxonomies and knowledge-centric concepts. Training was conducted in multiple stages, starting at a 512x512 resolution and fine-tuning at progressively higher resolutions up to 4K. The post-training phase is extensive, incorporating Continuing Training (CT) for foundational knowledge, Supervised Fine-Tuning (SFT) for artistic quality, and Reinforcement Learning from Human Feedback (RLHF) to align outputs with human preferences. A sophisticated Prompt Engineering (PE) module, built upon the Seed1.5-VL vision-language model, is used to process user inputs and enhance instruction following.
Intended Use & Applications
Seedream 4.5 is designed for professional creators and applications demanding high-quality, consistent, and controllable image generation. Its intended uses include:
- Professional Content Creation: Generating cinematic-quality visuals for digital advertising, social media, and print.
- Advanced Photo Editing: Performing complex edits, such as changing clothing materials, modifying backgrounds, or adjusting lighting, while maintaining subject integrity.
- E-commerce and Product Visualization: Creating high-quality product showcases and marketing materials.
- Graphic Design: Designing posters, key visuals, and other materials that require the integration of stylized text and typography.
- Creative Storytelling: Producing sequential, thematically related images for storyboards or visual narratives.
Performance
Seedream 4.5 and its predecessor, Seedream 4.0, have demonstrated top-tier performance on public benchmarks. The models are evaluated on the Artificial Analysis Arena, a real-time competitive leaderboard that ranks models based on blind user votes.
Text-to-Image Leaderboard (December 2025)
| Rank | Model | Developer | ELO Score | Release Date |
|---|---|---|---|---|
| 1 | GPT Image 1.5 (high) | OpenAI | 1,252 | Dec 2025 |
| 2 | Nano Banana Pro | 1,223 | Nov 2025 | |
| 5 | Seedream 4.0 | ByteDance Seed | 1,193 | Sept 2025 |
| 7 | Seedream 4.5 | ByteDance Seed | 1,169 | Dec 2025 |
