muapi-models

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Muapi models

Muapi 模型

Muapi is one API (

x-api-key

) in front of 500+ models: image, video, audio and speech, lipsync, 3D, LLM text, SEO/data tools, social publishing and utilities. Provider names are hidden; you only see Muapi model names, categories, schemas and costs. Model names are read live, so new models work immediately. Never rely on a remembered model list.

Workflow: discover -> inspect -> estimate -> run -> poll -> download.

Muapi 通过一个 API（

x-api-key

）接入 500 多个模型：图像、视频、音频和语音、唇形同步、3D、LLM 文本、SEO/数据工具、社交媒体发布和实用工具。提供商名称均被隐藏；你只会看到 Muapi 模型名称、类别、schema 和费用。模型名称为实时读取，因此新模型可立即使用。切勿依赖记忆中的模型列表。

工作流程：发现 -> 检查 -> 估算 -> 运行 -> 轮询 -> 下载。

Setup

设置

Create a key at https://muapi.ai/dashboard (or a free sandbox key, see below).

Export it and restart your agent from that same shell:

macOS/Linux:
```
export MUAPI_API_KEY=your_key
```

Windows PowerShell:

$env:MUAPI_API_KEY = "your_key"

and

setx MUAPI_API_KEY "your_key"

Keep the key in the environment only. Never write it into project files, commits or chat.

Sandbox key (returns mock outputs instantly, no credits spent), useful for testing a flow:

bash

curl -s -X POST https://api.muapi.ai/api/v1/keys -H "Content-Type: application/json" -d '{"is_test": true}'

Auth is always the

x-api-key: $MUAPI_API_KEY

header. If the

muapi

CLI is installed (

pip install muapi-cli

npm i -g muapi-cli

) prefer it, since it handles upload, polling and

--output-json

. Otherwise use the HTTP calls below (

curl

; on Windows PowerShell use

curl.exe

在 https://muapi.ai/dashboard 创建密钥（或使用下文的免费 sandbox 密钥）。
导出密钥，并从同一个 shell 重启你的 Agent：
- macOS/Linux：
```
export MUAPI_API_KEY=your_key
```
- Windows PowerShell：
```
$env:MUAPI_API_KEY = "your_key"
```
  ，并运行
```
setx MUAPI_API_KEY "your_key"
```
仅将密钥保存在环境变量中。切勿将其写入项目文件、提交记录或聊天内容。

Sandbox 密钥会立即返回模拟输出，不消耗额度，适合测试流程：

bash

curl -s -X POST https://api.muapi.ai/api/v1/keys -H "Content-Type: application/json" -d '{"is_test": true}'

身份验证始终使用

x-api-key: $MUAPI_API_KEY

请求头。如果已安装

muapi

CLI（

pip install muapi-cli

或

npm i -g muapi-cli

），建议优先使用它，因为它可以处理上传、轮询和

--output-json

。否则请使用以下 HTTP 调用（

curl

；在 Windows PowerShell 上使用

curl.exe

）。

Rules

规则

User instructions first, then any tool they already configured, then Muapi.
Always inspect before you run. Never guess fields or endpoint paths; schemas change.
Estimate cost before spending credits on anything non-trivial; check balance if no budget was given.
Submit, then poll every 3-10 seconds. Don't hold one long request open.
On
```
failed
```
, show the
```
error
```
field verbatim. Don't silently retry.
Name the model explicitly if the user did; otherwise state which model you picked and why.

优先遵循用户指令，然后使用用户已配置的任何工具，最后才使用 Muapi。
运行前始终进行检查。切勿猜测字段或 endpoint 路径；schema 会发生变化。
对任何非轻量级操作，在消耗额度前估算成本；如果未提供预算，则检查余额。
提交后，每 3～10 秒轮询一次。不要保持一个长时间请求一直开启。
遇到
```
failed
```
时，逐字显示
```
error
```
字段。不要静默重试。
如果用户指定了模型，请明确说出模型名称；否则请说明你选择了哪个模型以及原因。

1. Discover

1. 发现

bash

curl -s "https://api.muapi.ai/api/v1/models?category=Text%20to%20Image"
curl -s "https://api.muapi.ai/api/v1/models?family=seo"

There is no free-text search param: fetch a category and filter client-side by name/description, and try a few keyword variants. Categories:

Text to Image

Image to Image

Text to Video

Image to Video

Video to Video

Audio to Video

(lipsync),

Text to Audio

Image to 3D

Text to 3D

Text to Text

(LLMs and data tools),

Lora Support

Training

. CLI:

muapi discover "cinematic image to video" --output-json

bash

curl -s "https://api.muapi.ai/api/v1/models?category=Text%20to%20Image"
curl -s "https://api.muapi.ai/api/v1/models?family=seo"

没有自由文本搜索参数：获取一个类别，然后在客户端按名称或描述进行筛选，并尝试几种关键词变体。类别包括：

Text to Image

、

Image to Image

、

Text to Video

、

Image to Video

、

Video to Video

、

Audio to Video

（唇形同步）、

Text to Audio

、

Image to 3D

、

Text to 3D

、

Text to Text

（LLM 和数据工具）、

Lora Support

、

Training

。 CLI：

muapi discover "cinematic image to video" --output-json

2. Inspect

2. 检查

bash

curl -s "https://api.muapi.ai/api/v1/models/{name}"

Returns the input schema (required fields, allowed values, defaults), pricing and the real

endpoint

to POST to. CLI:

muapi inspect {name} --output-json

bash

curl -s "https://api.muapi.ai/api/v1/models/{name}"

返回输入 schema（必填字段、允许值、默认值）、定价以及要发送 POST 请求的实际

endpoint

。 CLI：

muapi inspect {name} --output-json

3. Estimate cost

3. 估算成本

bash

curl -s -X POST "https://api.muapi.ai/api/v1/models/{name}/estimate-cost" \
  -H "Content-Type: application/json" -d @payload.json

Same body you would send to run it. Balance:

GET /api/v1/account/balance

(CLI:

muapi account balance

bash

curl -s -X POST "https://api.muapi.ai/api/v1/models/{name}/estimate-cost" \
  -H "Content-Type: application/json" -d @payload.json

使用与运行模型时相同的请求体。余额查询：

GET /api/v1/account/balance

（CLI：

muapi account balance

）。

4. Upload input files

4. 上传输入文件

Models that take an image, video or audio need a hosted URL. Upload local files first:

bash

curl -s -X POST https://api.muapi.ai/api/v1/upload_file \
  -H "x-api-key: $MUAPI_API_KEY" -F "file=@./input.png"

Limits: images 10MB, videos 50MB. CLI:

muapi upload file ./input.png --output-json

接受图像、视频或音频的模型需要一个托管 URL。请先上传本地文件：

bash

curl -s -X POST https://api.muapi.ai/api/v1/upload_file \
  -H "x-api-key: $MUAPI_API_KEY" -F "file=@./input.png"

限制：图像 10MB，视频 50MB。 CLI：

muapi upload file ./input.png --output-json

5. Run

5. 运行

POST to the

endpoint

from step 2 (never a path you assumed):

bash

curl -s -X POST "https://api.muapi.ai{endpoint}" \
  -H "x-api-key: $MUAPI_API_KEY" -H "Content-Type: application/json" -d @payload.json

向第 2 步返回的

endpoint

发起 POST 请求（切勿使用自行推测的路径）：

bash

curl -s -X POST "https://api.muapi.ai{endpoint}" \
  -H "x-api-key: $MUAPI_API_KEY" -H "Content-Type: application/json" -d @payload.json

-> {"request_id": "...", "status": "processing"}


CLI: `muapi run {name} --input payload.json --no-wait --output-json`


CLI：`muapi run {name} --input payload.json --no-wait --output-json`

6. Poll

6. 轮询

bash

curl -s "https://api.muapi.ai/api/v1/predictions/{request_id}/result" -H "x-api-key: $MUAPI_API_KEY"

status

processing

completed

failed

. On

completed

outputs

is an array of URLs. CLI:

muapi runs get {request_id} --wait --output-json

bash

curl -s "https://api.muapi.ai/api/v1/predictions/{request_id}/result" -H "x-api-key: $MUAPI_API_KEY"

status

为

processing

、

completed

或

failed

。状态为

completed

时，

outputs

是一个 URL 数组。 CLI：

muapi runs get {request_id} --wait --output-json

7. Save results

7. 保存结果

Result URLs are kept for 30 days, then expire. As soon as a task completes, download every output to the location the user asked for (for example

./public/hero.png

) and tell them the path:

bash

curl -L -o ./public/hero.png "{output_url}"

If the user wants it in their app, write the integration code instead of just a file.

结果 URL 会保留 30 天，之后过期。任务完成后，应立即将每个输出下载到用户要求的位置（例如

./public/hero.png

），并告知用户文件路径：

bash

curl -L -o ./public/hero.png "{output_url}"

如果用户希望将结果集成到应用中，请编写相应的集成代码，而不要只保存文件。

Writing good requests

编写高质量请求

Ask for or infer: the output destination, a specific model if they have a preference, and key parameters (aspect ratio, resolution, duration). Examples:

"Make a 1:1 product photo of a white ceramic mug with Muapi, save to ./public/mug.png"
"Generate a 30 second lo-fi background track with Muapi"
"Turn README.md into a spoken intro with a Muapi speech model"

询问或推断以下信息：输出目的地、用户有偏好时的具体模型，以及关键参数（宽高比、分辨率、时长）。示例：

“使用 Muapi 制作一张 1:1 的白色陶瓷杯产品照片，保存到 ./public/mug.png”
“使用 Muapi 生成一段 30 秒的 lo-fi 背景曲目”
“使用 Muapi 语音模型将 README.md 转换为口播简介”

Troubleshooting

故障排除

Problem	Fix
`MUAPI_API_KEY` missing	Export it and restart the agent from the same terminal
401	Check the key at https://muapi.ai/dashboard; header must be `x-api-key`
Insufficient credits	Top up in the dashboard, then rerun
`failed` task	Show `error` , adjust parameters from the schema, retry once
Expired result link (older than 30 days)	Re-run, and save outputs locally next time
`curl` rejects args on Windows	Use `curl.exe`

问题	解决方法
缺少 `MUAPI_API_KEY`	导出密钥，并从同一个终端重启 Agent
401	检查 https://muapi.ai/dashboard 中的密钥；请求头必须为 `x-api-key`
额度不足	在 dashboard 中充值，然后重新运行
任务状态为 `failed`	显示 `error` ，根据 schema 调整参数后重试一次
结果链接已过期（超过 30 天）	重新运行，并在下次运行时将输出保存到本地
Windows 上 `curl` 拒绝参数	使用 `curl.exe`

Other interfaces

其他接口

OpenAPI: https://api.muapi.ai/openapi.json · MCP: https://muapi.ai/.well-known/mcp.json · Human docs: https://muapi.ai/agent-skills

OpenAPI：https://api.muapi.ai/openapi.json · MCP：https://muapi.ai/.well-known/mcp.json · 面向用户的文档：https://muapi.ai/agent-skills