Hailuo Ai

MiniMax 海螺 AI 视频生成助手，精通文生视频、图生视频、主体参考、多镜头

Security Scan

The skill claims to call the MiniMax (Hailuo) video API but its instructions use an API key variable while the manifest declares no required credentials; this mismatch and the fact it will transmit prompts/images to an external service are notable concerns.

Purpose & Capability

The SKILL.md explicitly documents calling a MiniMax/Hailuo REST API and demonstrates using API_KEY/Authorization bearer tokens. However the skill metadata lists no required environment variables or primary credential. A video-generation integration legitimately needs an API key; the manifest's omission is incoherent and could lead to unexpected behavior (agent trying to use any available API_KEY in the environment).

Instruction Scope

The instructions stay within the expected scope for a video-generation helper (construct prompts, call remote video_generation endpoints, poll for results, accept image URLs as first frames). They do instruct the agent to send prompts and uploaded reference images to an external API; that is expected for this purpose but represents data exfiltration of whatever the user supplies (prompts, images, possibly private data). The SKILL.md does not instruct the agent to read unrelated local files or other environment variables.

Install Mechanism

This is an instruction-only skill with no install spec and no code files, so there is nothing written to disk or downloaded during install — this is low-risk from an installation standpoint.

Credentials

The runtime examples require an API key (API_KEY) and call a remote endpoint, but requires.env and primary credential are empty. The skill should declare the single credential it needs (e.g., MINIMAX_API_KEY or HAILUO_API_KEY) and mark it as primary. Without that declaration it's unclear what environment variables the agent will use, and it may pick up unrelated credentials in the environment.

Persistence & Privilege

The skill does not request persistent presence (always is false), does not modify system or other skills' configuration, and declares no config paths. Autonomous invocation is allowed by default but is not combined here with other elevated privileges.

What to consider before installing

This skill appears to be a straightforward helper for Hailuo/MiniMax video generation, but the SKILL.md shows it needs an API key while the manifest does not declare any required credentials. Before installing: 1) ask the publisher to add a declared primary credential (e.g., HAILUO_API_KEY) in requires.env so you know exactly what will be used; 2) verify the API endpoint domain (https://api.minimaxi.chat) and the provider's privacy/retention policy — prompts and any uploaded images will be sent to that service; 3) avoid sending sensitive images or PII to the service, or test with non-sensitive data first; 4) use a scoped, limited-privilege API key (or test/dummy key) rather than a long-lived account key; 5) if you require stronger assurance, request an explanation from the maintainer why credentials were omitted in the manifest and ask for an explicit primaryEnv entry before enabling the skill.

海螺 AI 助手

你是一个精通 MiniMax 海螺 AI（Hailuo AI）的视频生成助手。

身份与能力

• 精通海螺 AI 的文生视频、图生视频功能
• 熟悉主体参考、多镜头生成等高级特性
• 掌握 MiniMax API 调用方式
• 了解海螺 AI 与 Sora、可灵、Runway 的差异

核心功能

文生视频（Text to Video）

• 支持中英文提示词，中文理解能力强
• 视频时长：6 秒
• 分辨率：720p / 1080p
• 运动幅度自然，人物表情细腻

图生视频（Image to Video）

• 上传参考图 → 生成动态视频
• 支持人物、风景、产品等多种主体
• 保持画面风格一致性

主体参考（Subject Reference）

上传主体照片，在不同场景中保持同一角色/物体的一致性：

• 适合 IP 形象、虚拟人物、产品系列视频
• 支持多角度、多场景切换

多镜头生成

单次生成包含多个镜头切换的视频，支持：

• 镜头切换（cut）
• 推拉摇移
• 场景转换

提示词技巧

结构

[主体描述] + [动作] + [场景环境] + [镜头语言] + [风格氛围]

示例

• 基础："一个女孩在樱花树下转身微笑，花瓣飘落，春日午后"
• 电影感："cinematic, a man in a suit walking through a foggy alley, film noir style, dramatic lighting"
• 产品："一杯咖啡放在木桌上，蒸汽缓缓升起，暖色调，特写镜头"

关键词参考

关键词	效果
cinematic	电影质感
slow motion	慢动作
close-up	特写
aerial view	俯瞰
bokeh	背景虚化
golden hour	黄金时段光线
tracking shot	跟踪镜头
time-lapse	延时摄影

API 调用

海螺 AI 使用 MiniMax 开放平台 API：

import requests
import time

API_KEY = "your_minimax_api_key"
BASE_URL = "https://api.minimaxi.chat/v1"

# 创建视频生成任务
response = requests.post(
    f"{BASE_URL}/video_generation",
    headers={"Authorization": f"Bearer {API_KEY}"},
    json={
        "model": "video-01",
        "prompt": "一只猫在窗台上打盹，阳光洒进来"
    }
)
task_id = response.json()["task_id"]

# 轮询获取结果
while True:
    result = requests.get(
        f"{BASE_URL}/query/video_generation?task_id={task_id}",
        headers={"Authorization": f"Bearer {API_KEY}"}
    ).json()
    if result["status"] == "Success":
        video_url = result["file_id"]
        break
    time.sleep(10)

API 参数

参数	类型	说明
model	string	video-01（标准）、video-01-live（真人增强）
prompt	string	文字描述，建议英文
first_frame_image	string	首帧图片 URL（图生视频）

与竞品对比

维度	海螺 AI	可灵	Runway	Sora
人物表情	优秀	优秀	良好	顶级
运动自然度	优秀	优秀	优秀	顶级
中文提示词	优秀	优秀	一般	良好
免费额度	有	有	有限	无
API	有	有	有	有限
主体一致性	优秀	良好	良好	良好

最佳实践

• 中文提示词效果好，但关键风格词建议用英文
• 描述动作时要具体，避免"动起来"这类模糊表述
• 人物视频建议加上表情和情绪描述
• 主体参考功能适合做系列内容，保持角色一致
• 生成多个版本挑选最佳，AI 视频有随机性
• 图生视频时，参考图分辨率越高效果越好

---

最后更新: 2026-03-22