Wan 3.0 Prime API MCP Server
Enables MCP-capable clients to generate higher-fidelity videos via Wan 3.0 Prime, supporting text-to-video, image-to-video, multimodal reference-to-video, and async task status polling.
README
Wan 3.0 Prime API — Python Wrapper
A focused Python SDK and MCP server for the Wan 3.0 Prime API on MuAPI — the higher-fidelity tier of Wan 3.0, targeting sharper detail and steadier motion at a higher per-second price. It supports text-to-video, image-to-video, multimodal reference-to-video (up to 10 reference images, 5 reference videos, and 5 reference audios), synchronized audio, file upload, and asynchronous job polling.
<p align="center"><a href="https://youtu.be/T4Oddw74O44"><img src="https://i.ytimg.com/vi/T4Oddw74O44/maxresdefault.jpg" width="720"></a></p> <p align="center"><a href="https://youtu.be/T4Oddw74O44"><b>▶ Watch: How to Access Wan 3.0 API - Best Uncensored Alternative to Seedance 2 </b></a></p>
Related Projects
- Wan 3.0 Prime on MuAPI — Model landing page for the Wan 3.0 and Wan 3.0 Prime video-generation tiers.
- Wan 3.0 Prime Text to Video playground — Try the model in the browser.
- MuAPI API reference — Endpoint and prediction lifecycle documentation.
- Wan-3.0-API — The standard-tier Wan 3.0 Python SDK and MCP server this repo mirrors.
- wan-3.0-comfyui — ComfyUI custom nodes for Wan 3.0 via MuAPI.
- awesome-ai-video-models — compare AI video models by API, price, and speed.
- Open-Generative-AI — open-source AI media studio for image and video workflows.
- Seedance-2-API — Python SDK for ByteDance Seedance video generation.
- Veo-4-API — Python SDK for Google Veo AI video generation.
- Flux-3-Dev-API — Python SDK for FLUX 3 image and video generation.
- Generative-Media-Skills — agent-ready skills for automated media workflows.
- muapi-cli — command-line access to MuAPI image, video, and audio models.
Install
git clone https://github.com/Anil-matcha/Wan-3.0-Prime-API.git
cd Wan-3.0-Prime-API
pip install -r requirements.txt
cp .env.example .env
Set MUAPI_API_KEY in .env. Set WAN_PRIME_API_BASE_URL only if you use a compatible provider other than the default MuAPI base URL.
Quick start
from wan_api import WanPrimeAPI
api = WanPrimeAPI()
job = api.text_to_video(
"A cinematic tracking shot of a red fox crossing a snowy forest at sunrise",
resolution="720p",
aspect_ratio="16:9",
duration=5,
)
result = api.wait_for_completion(job["request_id"])
print(result)
Image to video
job = api.image_to_video(
prompt="The subject turns toward camera as a gentle breeze moves their hair.",
image_url="https://example.com/reference.jpg",
aspect_ratio="9:16",
duration=5,
)
Reference to video
job = api.reference_to_video(
prompt="The person from the reference image walks into the room shown in the reference video.",
images_list=["https://example.com/character.jpg"],
videos_list=["https://example.com/room.mp4"],
resolution="720p",
duration=5,
)
API surface
| Method | Purpose |
|---|---|
text_to_video() |
Create a higher-fidelity video with synchronized audio from a text prompt. |
image_to_video() |
Animate a source image, with optional end-frame guidance via last_image. |
reference_to_video() |
Condition a video on up to 10 reference images, 5 reference videos, and 5 reference audios. |
upload_file() |
Upload a local reference file. |
get_result() / wait_for_completion() |
Retrieve an asynchronous job's output. |
Every method also accepts resolution (480p, 720p, 1080p), duration (2-30 seconds), thinking_mode (deeper reasoning for complex prompts), enable_audio, and seed.
Wan 3.0 Prime vs. standard Wan 3.0
Wan 3.0 Prime exposes the exact same request shape as standard Wan 3.0 — the same prompt, resolution, aspect_ratio, duration, thinking_mode, enable_audio, and seed parameters across text-to-video, image-to-video, and reference-to-video — just pointed at the wan3.0-prime-* endpoints instead of wan3.0-*. Use it for hero content, client deliverables, and campaign finals where the extra sharpness and motion consistency are worth the higher rate; use Wan-3.0-API for everyday generations.
Indicative MuAPI pricing (per second of output, billed by resolution and duration):
| Resolution | Wan 3.0 (standard) | Wan 3.0 Prime |
|---|---|---|
| 480p | $0.05/sec | $0.068/sec |
| 720p | $0.10/sec | $0.14/sec |
| 1080p | $0.20/sec | $0.28/sec |
MCP server
Expose Wan Prime tools to MCP-capable clients:
python mcp_server.py
The server provides text_to_video, image_to_video, reference_to_video, and get_task_status tools.
Endpoint compatibility
The client uses the wan3.0-prime-text-to-video, wan3.0-prime-image-to-video, and wan3.0-prime-reference-to-video paths beneath WAN_PRIME_API_BASE_URL. If your provider names its endpoints differently, pass that provider's compatible base URL or adapt the small client module before use.
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。