ai-cli

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

ai-cli

Generate text, images, and video from the terminal using AI models.

通过AI模型从终端生成文本、图片和视频。

When to Use

使用场景

Use when you need to:

Generate images from text prompts or existing images
Generate video from text prompts or images
Generate text (summaries, explanations, code reviews) from prompts or piped content
Compare outputs across multiple models side-by-side
Build composable media pipelines by chaining commands via stdin/stdout

适用于以下需求：

从文本提示词或现有图片生成图片
从文本提示词或图片生成视频
从提示词或管道输入内容生成文本（摘要、解释、代码评审）
同时对比多个模型的输出结果
通过标准输入/输出链式调用命令，构建可组合的媒体处理流程

Prerequisites

前置条件

Requires

AI_GATEWAY_API_KEY

or a provider-specific key (e.g.

OPENAI_API_KEY

) in the environment.

需要在环境变量中配置

AI_GATEWAY_API_KEY

或特定服务商的密钥（例如

OPENAI_API_KEY

）。

Commands

命令

bash

ai text "explain this code"              # generate text
ai image "a sunset over mountains"       # generate an image
ai video "a spinning triangle"           # generate a video
ai models --type image                   # list available models

bash

ai text "explain this code"              # generate text
ai image "a sunset over mountains"       # generate an image
ai video "a spinning triangle"           # generate a video
ai models --type image                   # list available models

Key Flags

关键参数

-m, --model <id>       Model ID (provider/name or short name), comma-separated for multi-model
-o, --output <path>    Output file or directory
-n, --count <n>        Number of generations per model
-q, --quiet            Suppress progress output
--json                 Output structured metadata as JSON (paths, timing, success/failure)

-m, --model <id>       Model ID (provider/name or short name), comma-separated for multi-model
-o, --output <path>    Output file or directory
-n, --count <n>        Number of generations per model
-q, --quiet            Suppress progress output
--json                 Output structured metadata as JSON (paths, timing, success/failure)

Piping Patterns

管道使用模式

Chain commands for agent workflows:

bash

undefined

通过链式命令构建Agent工作流：

bash

undefined

Pipe content in for summarization

管道输入内容进行摘要

cat file.txt | ai text "summarize this" git diff | ai text "write a commit message"

Image-to-video pipeline

图片转视频流程

ai image "a dragon" | ai video "animate this"

Image editing via stdin

通过标准输入编辑图片

cat photo.png | ai image "make it a watercolor"

undefined

cat photo.png | ai image "make it a watercolor"

undefined

Structured Output

结构化输出

Use

--json

to get machine-readable results:

bash

ai image "a sunset" --json

Returns:

json

{
  "elapsed_ms": 3420,
  "count": 1,
  "results": [
    {
      "index": 1,
      "model": "openai/gpt-image-2",
      "elapsed_ms": 3420,
      "success": true,
      "file": "/path/to/output.png"
    }
  ]
}

使用

--json

参数获取机器可读的结果：

bash

ai image "a sunset" --json

返回结果：

json

{
  "elapsed_ms": 3420,
  "count": 1,
  "results": [
    {
      "index": 1,
      "model": "openai/gpt-image-2",
      "elapsed_ms": 3420,
      "success": true,
      "file": "/path/to/output.png"
    }
  ]
}

Multi-Model Comparison

多模型对比

bash

ai image "a sunset" -m "openai/gpt-image-1,bfl/flux-2-pro,xai/grok-imagine-image"

bash

ai image "a sunset" -m "openai/gpt-image-1,bfl/flux-2-pro,xai/grok-imagine-image"

Output Behavior

输出行为

Interactive (TTY): saves to file, prints path to stderr
Piped (non-TTY): writes raw content to stdout for chaining
-o <dir>
: saves inside directory with auto-generated names

交互式（TTY）: 保存到文件，在标准错误输出中打印路径
管道模式（非TTY）: 将原始内容写入标准输出，用于链式调用
-o <dir>
: 将文件保存到指定目录，文件名自动生成

Timeouts

超时设置

text/image: 120 seconds
video: 300 seconds

文本/图片：120秒
视频：300秒

Exit Codes

退出码

```
0
```
— success
```
1
```
— all generations failed
```
2
```
— partial failure (some succeeded)

```
0
```
— 成功
```
1
```
— 所有生成任务失败
```
2
```
— 部分失败（部分任务成功）