# Seedream Image

> Seedream Image is an Agent Skill for documents published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Generate and edit AI images with Seedream (ByteDance) via AceDataCloud API. Use when creating images from text prompts, editing existing images with inpainting/outpainting, or working with ultra-high-resolution outputs. Supports Seedream 2.0, 2.1, 3.0, and 3.0 Turbo models.

- Skill name: `seedream-image`
- Category: [Documents](https://www.openparable.com/skill/documents/1)
- Source: https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/ai-ml/seedream-image-acedatacloud-skills-2/SKILL.md
- Repository: majiayu000/claude-skill-registry (498 GitHub stars when collected)

## SKILL.md

# Seedream Image Generation

Generate and edit AI images through AceDataCloud's Seedream (ByteDance) API.

## Authentication

```bash
export ACEDATACLOUD_API_TOKEN="your-token-here"
```

## Quick Start

```bash
curl -X POST https://api.acedata.cloud/seedream/images \
  -H "Authorization: Bearer $ACEDATACLOUD_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "a cyberpunk cat wearing VR goggles in a neon city", "model": "seedream-3.0"}'
```

## Models

| Model | Resolution | Speed | Best For |
|-------|-----------|-------|----------|
| `seedream-2.0` | Standard | Fast | Quick drafts |
| `seedream-2.1` | Standard | Fast | Improved quality over 2.0 |
| `seedream-3.0` | Up to 2048×2048 | Standard | High-quality generation (default) |
| `seedream-3.0-turbo` | Up to 2048×2048 | Faster | Speed-optimized with near-3.0 quality |

## Workflows

### 1. Text-to-Image

```json
POST /seedream/images
{
  "prompt": "a serene Japanese garden with cherry blossoms and a red bridge",
  "model": "seedream-3.0",
  "width": 1024,
  "height": 1024
}
```

### 2. Image Editing (Inpainting / Outpainting)

Edit regions of an existing image using a mask.

```json
POST /seedream/images/edit
{
  "prompt": "replace with a golden sunset sky",
  "image_url": "https://example.com/photo.jpg",
  "mask_url": "https://example.com/mask.png",
  "model": "seedream-3.0"
}
```

## Parameters

### Generation

| Parameter | Values | Description |
|-----------|--------|-------------|
| `model` | `"seedream-2.0"`, `"seedream-2.1"`, `"seedream-3.0"`, `"seedream-3.0-turbo"` | Model to use |
| `width` | 512 – 2048 | Image width in pixels |
| `height` | 512 – 2048 | Image height in pixels |
| `seed` | integer | Seed for reproducible results |

### Editing

| Parameter | Required | Description |
|-----------|----------|-------------|
| `image_url` | Yes | URL of the source image to edit |
| `mask_url` | Yes | URL of the binary mask (white = edit region) |
| `prompt` | Yes | Describe what to place in the masked area |

## MCP Server

```bash
pip install mcp-seedream
```

Or hosted: `https://seedream.mcp.acedata.cloud/mcp`

Key tools: `seedream_generate_image`, `seedream_edit_image`

## Gotchas

- Maximum resolution is **2048×2048** — exceed this and requests fail
- Mask images must be **binary** (black/white) with the same dimensions as the source image
- `seedream-3.0-turbo` offers a good speed/quality trade-off for iterative workflows
- Results return a direct image URL — no task polling needed for generation
- Editing always requires both `image_url` and `mask_url`

---

Source: https://www.openparable.com/skill/documents/seedream-image (OpenParable). Cite the canonical URL above.