Geek-skills-seedream-imagegen

Seedream 4.0 Image Generation Skill

This Skill provides a complete workflow for generating professional-grade AI images using the ByteDance Seedream 4.0 API.

Core Features

High-Resolution Generation: Supports 2K (2048x2048) and 4K (4096x4096) resolutions
Flexible Aspect Ratios: Supports multiple ratios such as 1:1, 16:9, 9:16, 4:3, 3:2, 21:9, etc.
Batch Generation: Generate up to 15 images in a single request
Reference Image Guidance: Supports up to 3 reference images to control style and composition
Excellent Text Rendering: Can generate clear text content in images
Sequential Batch Generation: Generate a series of images with consistent style

Workflow

1. Gather User Requirements

Before starting generation, be sure to ask the user for the following information:

Required Information

Image Description: What kind of image does the user want to generate?
- Ask: "Please describe the image content you want to generate"
- Tip: Encourage the user to provide a detailed description, including subject, style, atmosphere, etc.

Recommended Information to Ask

Size Requirements:
- Ask: "What size image do you need? (2K standard quality or 4K high-definition quality)"
- Default: 2K
- Note: 2K is suitable for most purposes, 4K is suitable for printing or scenarios requiring high clarity
Aspect Ratio:
- Ask: "What aspect ratio image do you need?"
- Common options and uses:
  - 1:1 (Square) - Social media, avatars
  - 16:9 (Horizontal widescreen) - Web banners, video thumbnails
  - 9:16 (Vertical) - Phone wallpapers, Stories
  - 4:3 (Standard horizontal) - Presentations, printing
- Default: 1:1
Quantity:
- Ask: "How many images do you need to generate? (1-15 images)"
- Default: 1
- Note: Generating multiple images provides more choices

2. Optimize the Prompt

Based on the user's description, optimize the generation prompt. Refer to references/prompt_engineering.md for detailed guidance.

Prompt Optimization Principles:

Use structured description: [Subject] + [Style] + [Details] + [Quality modifiers]
Add lighting and atmosphere descriptions
Include specific style keywords
For images that need text, clearly specify the text content and font style

Example Transformation:

User input: "A cat"

Optimized prompt:

"A fluffy orange tabby cat sitting on a wooden windowsill, 
golden hour lighting streaming through lace curtains, 
cozy home interior, warm color palette, soft focus background, 
professional photography, highly detailed fur texture"

3. Choose Appropriate Parameters

Choose parameters based on user needs and image type:

Purpose and Parameter Reference Table

Purpose	size	aspect_ratio	max_images
Social Media Images	2K	1:1 or 9:16	1-3
Web Banners	2K	16:9 or 21:9	1
Print Posters	4K	4:3 or custom	1-2
Product Images	2K	3:2 or 4:3	3-5
Concept Design	2K	16:9	5-10
Marketing Materials	2K-4K	Depending on the medium	3-5

4. Execute Image Generation

Use the scripts/generate_image.py script to generate images:

Command-line method:

python scripts/generate_image.py \
  --prompt "优化后的提示词" \
  --size 2K \
  --aspect-ratio 16:9 \
  --max-images 1 \
  --api-key YOUR_API_KEY \
  --output-dir ./outputs

Python code method:

from scripts.generate_image import SeedreamImageGenerator

generator = SeedreamImageGenerator(api_key="YOUR_API_KEY")

paths = generator.generate(
    prompt="优化后的提示词",
    size="2K",
    aspect_ratio="16:9",
    max_images=1,
    output_dir="./outputs"
)

print(f"生成的图像: {paths}")

5. Processing results

Images will be saved in the specified output directory (default ./outputs)
Filename format: seedream_YYYYMMDD_HHMMSS_提示词前缀.png
If multiple images are generated, numbered suffixes will be added automatically

6. Iterative optimization

If the user is not satisfied with the result:

Ask what specifically needs to be adjusted
Optimize the prompt based on feedback:
- Add more detailed descriptions
- Adjust style keywords
- Modify lighting or atmosphere descriptions
Consider adjusting technical parameters:
- Increase resolution (2K → 4K)
- Change aspect ratio
- Generate more variants to choose from

Advanced features

Use reference images

When a specific style or brand consistency is needed:

generator.generate(
    prompt="提示词",
    image_input=[
        "https://example.com/reference1.jpg",
        "https://example.com/reference2.jpg"
    ],
    size="2K",
    output_dir="./outputs"
)

Reference image uses:

Guide the artistic style
Maintain brand visual consistency
Provide composition reference

Sequential batch generation

Generate a series of images with a consistent style:

generator.generate(
    prompt="系列图像的统一描述",
    max_images=5,
    sequential=True,  # 启用顺序生成
    output_dir="./outputs"
)

Applicable scenarios:

Product series images
Storyboard
Brand asset kit
Social media series content

Custom size

For special size requirements:

generator.generate(
    prompt="提示词",
    size="custom",
    width=3000,
    height=2000,
    output_dir="./outputs"
)

API configuration

Set the API key

Method 1: Environment variable (recommended)

export SEGMIND_API_KEY="your_api_key_here"

Method 2: Pass directly

generator = SeedreamImageGenerator(api_key="your_api_key_here")

Get the API key

Visit Segmind
Register an account
Generate an API key in account settings

Reference documentation

When more detailed information is needed, refer to the following documentation:

API detailed description: references/api_reference.md
- Complete API parameter description
- Request examples
- Error handling
- Performance metrics
Prompt engineering: references/prompt_engineering.md
- Prompt structure guide
- Style keyword library
- Use case examples
- Best practices

Quick reference for common scenarios

Scenario 1: Social media images

generator.generate(
    prompt="优化的社交媒体内容描述",
    size="2K",
    aspect_ratio="1:1",
    max_images=3,  # 生成多个供选择
    output_dir="./outputs/social_media"
)

Scenario 2: Marketing posters

generator.generate(
    prompt="专业的营销海报描述,包含标题文字",
    size="4K",  # 高清晰度
    aspect_ratio="4:3",
    max_images=1,
    output_dir="./outputs/marketing"
)

Scenario 3: Product concept images

generator.generate(
    prompt="详细的产品描述和风格要求",
    size="2K",
    aspect_ratio="3:2",
    max_images=5,  # 多角度探索
    output_dir="./outputs/product_concepts"
)

Scenario 4: Brand series design

generator.generate(
    prompt="统一的品牌风格描述",
    size="2K",
    max_images=10,
    sequential=True,  # 保持一致性
    image_input=["品牌logo_url", "色卡_url"],
    output_dir="./outputs/brand_series"
)

Troubleshooting

Issue: API returns a 401 error

Check whether the API key is set correctly
Confirm that the API key is valid and has sufficient quota

Issue: The generated image quality is not ideal

Optimize the prompt and add more details
Try using a reference image
Increase the resolution to 4K
Generate multiple images for comparison and selection

Issue: Generation takes too long

4K images take longer (4-6 seconds)
Batch generation increases the total time
Check the network connection

Issue: Text in the image is unclear

Explicitly specify the font style in the prompt
Increase the resolution
Add keywords such as "high contrast", "bold typography"

Notes

API quota: Each generation consumes quota, so monitor the remaining quota
Generation time: about 1.8 seconds for 2K, about 4-6 seconds for 4K
Batch limit: up to 15 images per request
Reference image limit: up to 3 reference images
Prompt length: 200-300 words or fewer is recommended

Best practices

Always optimize prompts: do not directly use the user's brief description
Ask about the requirements clearly: understanding the image's purpose helps choose the correct parameters
Provide multiple options: for important projects, generate 3-5 for selection
Use reference images: use reference images when a specific style is needed
Iterative optimization: continuously improve prompts and parameters based on feedback
Save successful cases: record effective prompt templates for later use