🎬 Moody MiniMax H3 Video Prompt Skill Release

I'm releasing my personal MiniMax H3 video prompting skill, designed to help you turn simple ideas into detailed, cinematic video prompts.

When your AI-generated video looks weird, characters move unnaturally, the camera randomly jumps around, or the whole scene feels like a collection of disconnected clips, it usually comes down to two things:

1. 🚫 The video model has limitations.

2. ✍️ Your prompt writing sucks.

And honestly, a lot of the time, it's probably #2. 😅

People constantly complain:

❌ "The character keeps teleporting."
❌ "The camera changes angles for no reason."
❌ "The movements look unnatural."
❌ "The AI completely ignores my instructions."

But here's the thing: a video prompt is not just an image prompt with some movement added.

You're not describing a picture anymore. You're directing an entire sequence of events.

🛑 Stop writing video prompts like image prompts.

Humans tend to describe what they want to see, but often forget to explain how things actually happen.

Where does the character start? How does the movement begin? What happens next? Where is the camera? What should stay consistent?

These details matter.

Instead of trying to figure all of this out yourself, feed Moody MiniMax H3 Prompt Skill into your LLM / AI Agent, describe your video idea in normal language, and let AI handle the directing. 🤖🎬

🎥 Think of It as Your Personal AI Film Director

This Skill goes beyond basic prompt enhancement. It's designed to help structure the visual, temporal, and audio elements of AI-generated video.

🎞️ Natural Motion & Continuity

Characters shouldn't randomly teleport, change poses between shots, or suddenly appear somewhere else.

The Skill emphasizes continuous movement, logical action sequences, consistent positioning, and natural transitions.

📷 Cinematic Camera Direction

Camera movement should serve the scene, not randomly change angles every two seconds.

The Skill helps describe framing, camera movement, shot transitions, visual composition, and cinematic pacing.

🎭 Character Consistency & Reference Control

When working with reference images, videos, or other visual assets, the Skill helps identify what should remain consistent and what can change.

Characters, clothing, environments, props, and visual details are tracked throughout the sequence.

🎙️ Dialogue, Sound & Audio Direction

Video isn't just about visuals.

The Skill also structures dialogue, speaker identification, environmental sounds, audio references, and background music, helping create a more coherent audiovisual experience.

🧩 Structured Video Prompting

For complex reference-based tasks, the Skill organizes information into clearly defined sections, including subject definitions, reference relationships, continuity analysis, detailed shot descriptions, and audio direction.

This is especially useful when combining multiple references into a single video.

💡 Less Random AI Chaos. More Actual Directing.

One of the biggest problems with AI video prompting is that people keep asking for cinematic results without actually describing the cinematography.

A good video prompt needs more than a beautiful scene. It needs direction, timing, movement, and continuity.

This Skill incorporates structured video-prompting techniques and cinematic principles to help your LLM translate ideas into clearer instructions for MiniMax H3.

You focus on the story and creativity. Let AI figure out how to communicate the scene.

AI knows how to communicate with AI better than most humans do.

💡 Before blaming the video model, fix the prompt.

❤️ Special Thanks

Special thanks to @noway163 for his contributions and inspiration. @lenfari for techinical support.

He's an absolute master of prompting and one of the most impressive AIGC "film directors" I've seen. 🎬🔥

You can find him in the Moody Discord channel.

If you're interested in cinematic prompting, camera language, shot composition, visual storytelling, and getting AI to actually understand your scenes, definitely check out his work.

🎬 Moody MiniMax H3 视频提示词 Skill 正式发布

这次我把自己的 MiniMax H3 视频提示词 Skill 也放出来了，专门用来帮助大家把简单的视频创意转换成更加完整、专业的电影感视频提示词。

当你发现 AI 生成的视频动作奇怪、人物莫名其妙地瞬移、镜头乱切，或者整段视频看起来像几个毫无关联的片段拼在一起时，通常就两个原因：

1. 🚫 视频模型本身的能力有限

2. ✍️ 你的提示词写得不行

说实话，很多时候，大概率是第二种。 😅

我经常看到有人吐槽：

❌ “人物怎么突然瞬移了？”
❌ “镜头为什么一直乱切？”
❌ “人物动作一点都不自然。”
❌ “AI 根本不按照我的要求来。”

但问题是，视频提示词并不是给图片提示词加上几个动作描述那么简单。

你现在不是在描述一张静态图片，而是在指导一整段视频应该如何发生。

🛑 别再用写图片提示词的方式写视频了。

很多人在写视频提示词的时候，只会描述自己想看到什么，却忽略了更重要的问题：

这个画面究竟应该怎样发生？

人物从哪里开始移动？动作怎么衔接？镜头应该在哪里？什么时候需要切镜头？哪些东西必须保持一致？

这些细节都会影响最终的视频效果。

所以，与其自己费尽心思研究怎么写，不如直接把 Moody MiniMax H3 Prompt Skill 喂给你的 LLM / AI Agent。

然后像平时聊天一样，告诉 AI 你想拍什么，让 AI 帮你完成专业的视频提示词编写。🤖🎬

🎥 相当于给你配了一个 AI 视频导演

这次的 Skill 不只是简单地帮你扩写提示词，而是针对 AI 视频生成的特点，加入了更加完整的视觉、时间和声音描述逻辑。

🎞️ 自然动作与镜头连续性

人物不应该上一秒还站着，下一秒突然换了姿势或瞬移到另一个地方。

Skill 会更加注重人物动作的自然衔接、空间关系、位置变化，以及镜头之间的连续性。

📷 电影级镜头语言

镜头不是切得越多就越有电影感。

什么时候应该固定机位？什么时候应该跟拍？什么时候适合推进镜头？什么时候才真正需要切换视角？

Skill 会帮助 AI 更合理地组织镜头运动、画面构图、景别变化以及整体节奏。

🎭 人物一致性与参考素材控制

当你使用参考图片、参考视频或其他素材时，Skill 会帮助 AI 区分哪些内容需要保留，哪些内容可以修改。

包括人物外观、服装、环境、道具以及不同镜头之间的视觉连续性。

🎙️ 对白、环境音与声音设计

视频不只有画面，声音同样重要。

Skill 包含了对白、说话者身份、环境音效、参考音频，以及背景音乐等内容的组织规则。

尤其是在多人对话的场景中，明确谁在说话、什么时候说话，以及声音和动作之间的关系，可以让提示词更加清晰。

🧩 结构化视频提示词

对于复杂的多参考素材任务，Skill 会将内容整理成更加清晰的结构，包括：

人物与素材定义、参考关系、内容保留分析、详细镜头描述、声音设计等。

让 AI 不只是理解你想要的画面，还能更清楚地理解不同素材之间应该如何组合。

💡 少一点 AI 乱来，多一点真正的导演思维。

很多人要求 AI 生成电影级画面，但写出来的提示词却只有一句：

“一个女孩走在街上，电影感，高清，专业摄影。”

然后就开始抱怨 AI 不听话。😂

电影感不是多写几个 Cinematic、Masterpiece、8K 就能解决的。

真正重要的是镜头语言、动作设计、时间逻辑、空间连续性，以及画面应该如何一步步发展。

这个 Skill 就是为了帮助你的 LLM 更合理地组织这些内容，把简单的创意转换成更加清晰、适合 MiniMax H3 使用的视频提示词。

你负责想故事，AI 负责当导演。

让 AI 去跟 AI 沟通，很多时候比人类自己硬写效果更好。

💡 所以在怪视频模型之前，先看看是不是你的提示词有问题。

❤️ 特别感谢

特别感谢 @noway163 的贡献与启发。@lenfari 的技术支持！

他是真的 Prompt 大师，也是我见过最强悍的 AIGC “导演”之一。🎬🔥

你可以在 Moody Discord 频道 找到他。

如果你对电影感提示词、镜头语言、构图、场景调度、人物动作、叙事，以及怎么让 AI 真正理解你脑子里的画面感兴趣，强烈建议去看看他的作品。