byted-seedream-image-generate — Claude Code skill
byted-seedream-image-generate generates images from text prompts via Volcano Engine Seedream. Choose from three model versions, batch generation, style transfer, and web search integration (5.0-lite).
Apache-2.0 304 installs ★ 494 updated 2026-10-22
npx skills add bytedance/agentkit-samples --skill byted-seedream-image-generate -a claude-codeWhat byted-seedream-image-generate does
This skill connects to three Seedream model versions. Version 4.0 prioritizes speed for everyday use. Version 4.5 adds detail for complex scenes. Version 5.0-lite delivers highest quality, supports web search to find and visualize real-world objects, custom output formats (PNG or JPEG), and best results for creative work.
Capabilities cover text-to-image generation, style transfer from reference images, batch creation up to 15 images, watermark control, and custom dimensions. Requires a Volcano Engine API key in environment variable ARK_API_KEY, MODEL_IMAGE_API_KEY, or MODEL_AGENT_API_KEY. Default model is 5.0, switchable via command flag. Output as URL or base64. Request timeout is 1200 seconds.
What it helps with
- generate artwork from a detailed description
- transform a photo into a different artistic style
- create multiple image variations in one batch
- produce high-fidelity visuals for design work
- visualize current objects found via web search
Example prompts for the agent
- Draw a kitten playing in a garden, watercolor style
- Convert this photo to anime art using version 5.0
- Generate three sunset landscape variations for a book cover
How to install byted-seedream-image-generate
1Skills CLI (Vercel)
npx skills add bytedance/agentkit-samples --skill byted-seedream-image-generate -a claude-code2Manually: put the files in place
~/.claude/skills/byted-seedream-image-generate/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/byted-seedream-image-generate/
1Skills CLI (Vercel)
npx skills add bytedance/agentkit-samples --skill byted-seedream-image-generate -a codex2Manually: put the files in place
.agents/skills/byted-seedream-image-generate/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add bytedance/agentkit-samples --skill byted-seedream-image-generate -a cursor2Manually: put the files in place
.cursor/skills/byted-seedream-image-generate/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install bytedance/agentkit-samples skills/byted-seedream-image-generate2Manually: put the files in place
.github/skills/byted-seedream-image-generate/SKILL.md1Skills CLI (Vercel)
npx skills add bytedance/agentkit-samples --skill byted-seedream-image-generate -a gemini-cli2Platform CLI
gemini skills install https://github.com/bytedance/agentkit-samples --path skills/byted-seedream-image-generateManually: put the files in place
.devin/skills/byted-seedream-image-generate/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add bytedance/agentkit-samples --skill byted-seedream-image-generate -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about byted-seedream-image-generate
How do I install byted-seedream-image-generate in Claude Code?
Run npx skills add bytedance/agentkit-samples --skill byted-seedream-image-generate -a claude-code, then check ~/.claude/skills/byted-seedream-image-generate/SKILL.md for setup.
What does byted-seedream-image-generate do?
Creates and transforms images using Seedream models. Generates pictures from descriptions, applies style transfer, runs batch jobs, and integrates web search to find and visualize real-world subjects (5.0 only).
Which model version should I use?
Start with 5.0-lite—it's the highest quality and most versatile. Pick 4.0 for speed, 4.5 for detail-rich scenes. Web search and custom formats only work in 5.0.
Source text
SKILL.md · Apache-2.0
---
name: byted-seedream-image-generate
description: Generate high-quality images from text prompts using Volcano Engine Seedream models. Supports multiple artistic styles and aspect ratios. Use this skill when users want to create images from text descriptions, generate artwork in various styles, create visual content for creative projects, or need AI-powered image generation capabilities.
license: Apache-2.0
tags: ["image-generation", "seedream", "volcengine", "ai-art", "text-to-image", "image-to-image"]
version: 1.0.0
---
# byted-seedream-image-generate
## Description
Generate high-quality images from text prompts using Volcano Engine Seedream models. This skill provides access to three powerful Seedream model versions (4.0, 4.5, and 5.0-lite), each offering unique capabilities for different use cases.
## When to Use This Skill
Use this skill when:
- Users want to create images from text descriptions
- Users need to generate artwork in various artistic styles
- Users want to create visual content for creative projects
- Users need AI-powered image generation capabilities
- Users want to convert reference images to different styles
- Users need to generate multiple images in batch
- Users require high-quality, professional-looking images
## Model Versions
| Version | Model Name | Release Date | Recommendation | Best For |
|---------|------------|--------------|----------------|----------|
| 4.0 | doubao-seedream-4-0-250828 | August 2025 | ⭐⭐⭐ | Daily use, quick generation |
| 4.5 | doubao-seedream-4-5-251128 | November 2025 | ⭐⭐⭐⭐ | Detail-oriented work, complex scenes |
| 5.0 | doubao-seedream-5-0-260128 | 2026 | ⭐⭐⭐⭐⭐ | Highest quality, best creativity, tools support |
## Features
- **Text-to-Image**: Generate images from detailed text descriptions
- **Image-to-Image**: Transform reference images into different styles
- **Batch Generation**: Create multiple images in a single request
- **Multiple Versions**: Choose from 4.0, 4.5, or 5.0-lite models
- **Watermark Control**: Option to disable watermarks
- **Custom Sizes**: Support for various image dimensions
- **Output Formats**: PNG and JPEG formats (5.0-lite only)
- **Web Search Tools**: Internet search integration (5.0-lite only)
## Installation & Setup
### Prerequisites
```bash
# Required: API Key configuration
export MODEL_IMAGE_API_KEY="your-api-key-here"
# or
export MODEL_AGENT_API_KEY="your-api-key-here"
# or
export ARK_API_KEY="your-api-key-here"
# Optional: API Base URL (default already configured)
export MODEL_IMAGE_API_BASE="https://ark.cn-beijing.volces.com/api/v3"
# or
export ARK_BASE_URL="https://ark.cn-beijing.volces.com/api/v3"
```
The script will prioritize:
1. Environment variables (`ARK_API_KEY`, `MODEL_IMAGE_API_KEY`, `MODEL_AGENT_API_KEY`)
2. Environment variables (`ARK_BASE_URL`, `MODEL_IMAGE_API_BASE`)
3. Default values
## Usage
### Basic Usage (5.0-lite version recommended)
```bash
cd scripts
python seedream_image_generate.py -p "A cute kitten playing in a garden"
```
### Specify Version
```bash
# Use 4.0 version
python seedream_image_generate.py -p "A cute kitten" --version 4.0
# Use 4.5 version
python seedream_image_generate.py -p "A cute kitten" --version 4.5
# Use 5.0-lite version (recommended)
python seedream_image_generate.py -p "A cute kitten" --version 5.0
```
### Advanced Options
```bash
# Custom size without watermark
python seedream_image_generate.py -p "Beautiful sunset" -s 2048x2048 --no-watermark --version 5.0
# Batch generation
python seedream_image_generate.py -p "Generate 3 cute dog pictures" -g --max-images 3 --version 4.5
# Image-to-image
python seedream_image_generate.py -p "Convert this image to anime style" -i "https://example.com/image.jpg" --version 5.0
# Web search tool (5.0-lite only)
python seedream_image_generate.py -p "Latest 2026 smartphone" --web-search --version 5.0
# Custom output format (5.0-lite only)
python seedream_image_generate.py -p "A beautiful landscape" --output-format png --version 5.0
# List all supported versions
python seedream_image_generate.py --list-versions
```
## Command Line Options
| Option | Shortcut | Description | Default |
|--------|----------|-------------|---------|
| `--prompt` | `-p` | Image description text (required) | - |
| `--version` | `-v` | Version selection: `4.0`, `4.5`, `5.0` | `5.0` |
| `--size` | `-s` | Image dimensions | `2048x2048` |
| `--image` | `-i` | Single reference image URL | - |
| `--images` | - | Multiple reference image URLs (space separated) | - |
| `--group` | `-g` | Enable batch image generation | `false` |
| `--max-images` | - | Maximum images for batch generation | `15` |
| `--output-format` | - | Output format: `png` or `jpeg` (5.0 only) | `jpeg` |
| `--response-format` | - | Response format: `url` or `b64_json` | `url` |
| `--stream` | - | Enable streaming output | `false` |
| `--web-search` | - | Enable web search tool (5.0 only) | `false` |
| `--optimize-prompt-mode` | - | Prompt optimization mode: `standard` or `fast` | - |
| `--timeout` | `-t` | Timeout in seconds | `1200` |
| `--no-watermark` | - | Disable watermark | `false` |
| `--list-versions` | - | List all supported versions | - |
## Python API Usage
```python
import asyncio
import sys
sys.path.append("scripts")
from seedream_image_generate import seedream_generate
async def main():
# Use 5.0-lite version (default)
result = await seedream_generate([
{
"prompt": "A cute kitten",
"size": "2048x2048",
"watermark": False,
"output_format": "png" # 5.0-lite only
}
], version="5.0")
print(result)
asyncio.run(main())
```
## Version Selection Guide
### Choose 4.0 if:
- You need quick daily generation
- Quality requirements are not extremely high
- You need faster generation speed
- Simple scenes and styles
### Choose 4.5 if:
- You want richer details
- You're working with complex scenes
- You need better style reproduction
- You have moderate quality requirements
### Choose 5.0-lite (Recommended) if:
- You want the highest quality
- You need breakthrough creative expression
- You have extreme detail requirements
- **You need tools parameter (like web search)** ⭐
- **You need custom output format (png/jpeg)** ⭐
- Important projects and work
**When in doubt, use 5.0-lite!** ⭐
## Prompt Engineering Tips
### Basic Prompt Structure
```
[Subject Description] + [Style/Art Movement] + [Lighting/Atmosphere] + [Quality/Resolution]
```
### Advanced Prompts (Optimized for 5.0-lite)
```
[Subject Description], [Creative Style/Art Movement], [Unique Perspective/Composition], [Special Lighting/Atmosphere], [Emphasizing 5.0-lite creative expression]
```
## Parameter Support by Version
| Parameter | Seedream 4.0 | Seedream 4.5 | Seedream 5.0-lite | Description |
|-----------|--------------|--------------|-------------------|-------------|
| model | ✅ | ✅ | ✅ | Model name |
| prompt | ✅ | ✅ | ✅ | Prompt (required) |
| image | ✅ | ✅ | ✅ | Reference image(s) |
| size | ✅ | ✅ | ✅ | Image dimensions |
| sequential_image_generation | ✅ | ✅ | ✅ | Batch generation control |
| sequential_image_generation_options | ✅ | ✅ | ✅ | Batch generation config |
| response_format | ✅ | ✅ | ✅ | Response format (url/b64_json) |
| watermark | ✅ | ✅ | ✅ | Watermark (true/false) |
| stream | ✅ | ✅ | ✅ | Streaming output |
| optimize_prompt_options | ✅ | ✅ | ✅ | Prompt optimization config |
| **tools** | ❌ | ❌ | **✅** | Tool configuration (5.0-lite only) |
| **output_format** | ❌ | ❌ | **✅** | Output format (png/jpeg, 5.0-lite only) |
## Final Return Info
### For OpenClaw
1. **First, save/download the generated image to an allowed directory**:
- Recommended: `~/.openclaw/workspace/`
2. **Use the `message tool` to send the image** with these parameters:
```json
{
"action": "send",
"message": "Optional text description",
"media": "/root/.openclaw/workspace/generated-image.jpg"
}
```
- The tool will automatically: read the local file, upload to the channel server as a web URL, and send as an image message
3. **Verify success**: Check that the tool returns `{"ok": true}` to confirm the image was sent successfully
4. **Normal Text** not message tool: After generation, show list of images with Markdown format, for example:

**Very important**:
The image URL must be an **original online resource link** starting with `https://`, **not** a local image address you have downloaded.
### For Normal
You should return three types of information:
1. File format, return the image file (if you have some other methods to send the image file) and the local path of the image, for example:
local_path: /root/.openclaw/workspace/skills/image-generate/xxx.png
2. After generation, show list of images with Markdown format, for example:
```


```
## FAQ
### Q: What's the difference between the versions?
A: 4.0 is for quick daily use, 4.5 offers better details for complex scenes, and 5.0-lite provides the highest quality with unique tools support.
### Q: How long are generated URLs valid?
A: URLs are valid for 24 hours. Please download and save your images promptly.
### Q: What image formats are supported for references?
A: Common formats like JPG and PNG are supported, provided via URL or Base64.
### Q: Can I use multiple versions in one call?
A: Currently, only one version per call. For comparisons, make separate calls for different versions.
## License
This skill is licensed under the Apache License 2.0. See the LICENSE file for details.
## Notice
Please comply with Volcano Engine's terms of service and relevant laws and regulations when using this skill.
© bytedance · Apache-2.0 · bytedance/agentkit-samples