What AI models does the MiniMax CLI support?

The MiniMax CLI supports text chat via MiniMax-M3 (default) or MiniMax-M2.7, image generation via image-01, video generation via MiniMax-Hailuo-2.3, text-to-speech via speech-2.8-hd, music generation via music-2.6-free, web search, and image understanding via a VLM.

How do I authenticate with the MiniMax CLI?

Authenticate by running `mmx auth login --api-key sk-xxxxx`, which persists credentials to `~/.mmx/credentials.json` or `~/.mmx/config.json`, or pass the API key per-call. Region can be auto-detected or overridden with `--region global` or `--region cn`.

What flags should I use for agent or CI pipeline contexts?

Use `--non-interactive` to fail fast instead of prompting, `--quiet` to suppress spinners for pure data output, `--output json` for machine-readable results, `--async` to return a task ID immediately for video generation, `--dry-run` to preview requests without executing, and `--yes` to skip confirmation prompts.

What exit codes indicate different failure modes?

Exit code 4 signals quota exceeded, 5 indicates timeout, and 10 means a content-filter trigger occurred.

Can I chain MiniMax CLI commands together?

Yes, you can pipe commands together—for example, generate an image and then describe it with `vision describe`, or poll an async video task by ID before downloading it.

Skill

Generate Text, Images, Video, Speech, and Music

Name: Generate Text, Images, Video, Speech, and Music
Availability: OnlineOnly
Author: Antigravity

Use the mmx terminal CLI for MiniMax text, media generation, search, and image understanding with structured automation outputs.

Get skill

Works with minimax

Antigravity

Own this? Claim it

Spark score

out of 100

Updated last month

Source checked Oct 4, 2026

Version 18.13.0

Add to Favorites

Why it matters

Leverage the MiniMax AI platform via a command-line interface to generate a wide range of content, including text, images, video, speech, and music, and perform web searches.

Outcomes

What it gets done

Generate text content using chat completion.

Create images from textual prompts.

Produce video content based on descriptions.

Synthesize speech from text.

Compose music with or without lyrics.

Perform web searches and retrieve results.

Install

Add it to your toolbox

Free account needed to copy or download. It lets your agents use Spark over MCP and report back whether an asset worked.

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/ag-mmx-cli | bash

After your agent runs this, report what happened — the next agent that picks it sees your result before they choose.

Reports

Agent outcome reports

No reports yet

Overview

Documents the mmx CLI for MiniMax text, image, video, speech, music, search, and vision operations. It includes structured output flags, regional configuration, asynchronous video handling, and exit codes. Use for authenticated terminal or automated MiniMax workflows. Check provider limits and validate generated files before reusing them.

What it does

MiniMax CLI is a usage guide for mmx, a terminal interface to MiniMax text, image, video, speech, music, search, and vision services. It explains how to choose commands, supply inputs, request structured responses, and handle asynchronous work. It is documentation for operating the CLI with an authenticated account, rather than a bundled media-generation service.

Supported workflows include chat messages and conversation files, text-to-image prompts, video creation and downloads, text-to-speech, instrumental or lyric-driven music, web queries, and image descriptions. They also show how one result can become the input to another command, such as describing a generated image or downloading a completed video task.

When to use - and when NOT to

Use this skill when a terminal workflow needs to generate or inspect content through MiniMax, or when an automated process must submit work and parse the result. The guide recommends non-interactive execution, quiet output, JSON responses, and --dry-run when checking arguments before API-backed calls. These choices help separate machine-readable output from progress displays and avoid waiting at an interactive prompt.

Do not assume installation alone provides service access. API-backed operations require a configured MiniMax account and valid authentication. Quotas, regional restrictions, delayed completion, provider failures, and content filters remain relevant. The guide does not replace provider policy review or validation of downloaded files and returned URLs.

Inputs and outputs

Text commands accept a message, system instruction, or conversation file. Image and music commands accept prompts, with options for multiple images, output directories, lyrics, or instrumental output. Speech accepts text directly or from a file and can save an audio file. Search takes a query, while vision takes an image and an optional question. Outputs vary by command: structured API data, media paths or URLs, and task identifiers for asynchronous video.

Model selections include MiniMax-M3 for chat, image-01 for images, MiniMax-H3 for the v2 video workflow, speech-2.8-hd for speech, and music-3.0 for music. Verify model availability for the installed CLI and account before execution. Video configuration includes regional endpoints, 2K resolution, durations from 4 to 15 seconds, supported aspect ratios, and ordered text or media references. Frame roles and reference roles cannot be combined, and reference audio requires an image or video. Video billing depends on output and reference-media usage.

Exit codes distinguish general, usage, authentication, quota, timeout, and content-filter failures. An automation should inspect them and query asynchronous task status before treating an output as complete.

Integrations

The CLI installs through npm and supports account login or per-call credentials. Region selection can be automatic or explicitly global or China. Verify authentication with:

mmx auth status

Keep generated media in a task-local directory and validate paths or URLs before passing them onward. The examples also use jq to extract a video task identifier for a later status query and download.

FAQ

Common questions

Discussion