使用说明
duomi-image-video-mcp
MCP server for DuomiAPI — AI image generation, image editing, and video generation.
Tools
| Tool | Description |
|---|---|
generate_image |
Text → image (nano-banana) — returns task_id |
edit_image |
Image + prompt → edited image (nano-banana) — returns task_id |
get_image_task |
Poll image task status — returns state + image URLs |
generate_video |
Text/image → video (VEO) — returns task_id |
get_video_task |
Poll video task status — returns state + progress + video URLs |
Setup
Get an API key from duomiapi.com.
Claude Code
Add to ~/.claude.json:
{
"mcpServers": {
"duomi-image-video": {
"command": "npx",
"args": ["-y", "duomi-image-video-mcp"],
"env": {
"DUOMI_API_KEY": "your_key_here"
}
}
}
}
CLI
DUOMI_API_KEY=your_key npx duomi-image-video-mcp
# or
npx duomi-image-video-mcp --api-key your_key
Models
Image (nano-banana):
| Model ID | Alias | Notes |
|---|---|---|
gemini-3.1-flash-image-preview |
nano-banana-2 | Fast, supports 2K/4K |
gemini-2.5-flash-image |
nano-banana | Standard |
gemini-3-pro-image-preview |
nano-banana-pro | Highest quality |
Video (VEO):
| Model ID | Notes |
|---|---|
veo3.1-fast |
Faster, lower cost |
veo3.1-pro |
Higher quality |
Video Generation Types
| Type | Description |
|---|---|
TEXT |
Text-to-video (default) |
FIRST&LAST |
Use 1–2 images as first/last frame |
REFERENCE |
Use up to 3 images as visual references (16:9 only) |
Workflow
generate_image(prompt) → { task_id }
→ get_image_task(task_id) every 5s → { state: "succeeded", images: [{ url }] }
→ edit_image(prompt, image_urls) → { task_id }
→ get_image_task(task_id) every 5s → { state: "succeeded", images: [{ url }] }
generate_video(prompt) → { task_id }
→ get_video_task(task_id) every 10s → { state: "succeeded", progress: 100, videos: [{ url }] }
License
MIT
