| name | h3-prompt-writing |
|---|---|
| description | Write MiniMax H3 video generation prompts for T2VA, I2VA, FL2VA, L2VA, and Ref2VA. 为 MiniMax H3 的 T2VA、I2VA、FL2VA、L2VA 和 Ref2VA 编写视频生成提示词。 Use when rewriting multimodal requests into H3 prompt structures, composing integrated_multimodal_description, overall_soundscape, and non_diegetic_music, aligning keyframes, or defining reference labels for images, videos, and audio. 适用于将多模态生成需求改写为 H3 提示词结构，编写 integrated_multimodal_description（综合多模态描述）、overall_soundscape（整体声景） 和 non_diegetic_music（非画面内音乐），aligning keyframes（对齐关键帧），以及为图像、视频和音频等参考素材定义引用标签。 |
| compatibility | Portable to any agent that can read local files — no external API calls, MiniMax Hub tools, or proprietary runtime required. 可移植到任何能够读取本地文件的 Agent——无需调用外部 API、MiniMax Hub 工具或专有运行时。 The agents/openai.yaml file only adds optional ChatGPT/Codex UI metadata; it does not restrict the skill to OpenAI agents. agents/openai.yaml 文件仅用于添加可选的 ChatGPT/Codex UI 元数据，并不会将该 Skill 限制为只能由 OpenAI Agent 使用。 |
Workflow 工作流
-
Identify the input mode: T2VA, I2VA, FL2VA, L2VA, or full-reference Ref2VA.
确定输入模式：T2VA、I2VA、FL2VA、L2VA，或全参考模式 Ref2VA。 -
For base text/keyframe modes, read
对于基础文本/关键帧模式，读取 references/base-en.txt，并遵循其中定义的最终提示词结构。references/base-en.txtand follow its final prompt structure. -
For full-reference mode, read
对于全参考模式，读取 references/ref-en.txt，并遵循其中定义的六部分改写格式。references/ref-en.txtand follow its six-section rewrite format. -
Preserve the exact field names, section order, labels, and timing notation from the selected guide.
严格保留所选指南中的字段名称、章节顺序、标签和时间标记格式。
Base Modes 基础模式
-
T2VA: build the full audiovisual timeline from text.
根据文本构建完整的视听时间线。 -
I2VA: start from the first frame and develop forward from it.
以首帧为起点，并在此基础上向后续内容展开。 -
FL2VA: describe the continuous path between the first and last frames.
描述从首帧到尾帧之间连续、完整的演变过程。 -
L2VA: infer a plausible opening and converge to the supplied last frame.
推断一个合理的起始场景，并逐步过渡、收敛到所提供的尾帧。
Use integrated_multimodal_description, overall_soundscape, and non_diegetic_music in the order shown in references/base-en.txt.
Full-Reference Mode 全参考模式
Ref2VA rewrites use subject_definitions, summary, retention_analysis, detailed_description, overall_soundscape, and non_diegetic_music in that order.
Ref2VA 的提示词改写按照以下顺序组织：subject_definitions（主体定义）、summary（总结）、retention_analysis（保留分析）、detailed_description（详细描述）、overall_soundscape（整体声景） 和 non_diegetic_music（非画面内音乐）。
Reference labels stay consistent across all sections.
所有部分中的参考素材标签必须保持一致。Read references/ref-en.txt for label rules, retention analysis, and complete examples.
Output Rules 输出规则
-
Write rewrite sections in English; preserve dialogue, lyrics, and visible scene text in their original language.
改写后的各部分使用英文编写；对话、歌词以及场景中可见的文字保留其原始语言。 -
Describe each shot by composition, subjects, environment, actions, camera, sound, and the exact point where referenced content appears.
描述每个镜头时，应包含构图、主体、环境、动作、镜头运动、声音，以及参考内容出现的准确时间点。 -
Avoid plot summaries, unresolved reference labels, and timing that does not match the requested duration.
避免仅进行剧情概述、使用未明确对应关系的参考标签，以及使用与所请求视频时长不匹配的时间标记。
Tips for Better Results 获得更好生成效果的技巧
-
Always match the total duration of the description to the requested video length (4–15 seconds).
始终确保描述的总时长与请求生成的视频时长一致（4–15 秒）。 -
Keep reference labels consistent (e.g.
在所有部分中保持参考素材标签一致，例如<Picture 1>,<Video 1>,<Audio 1>) across every section.<Picture 1>、<Video 1>、<Audio 1>。 -
Prefer concrete visual and audio details over abstract words like "cinematic" or "beautiful".
优先使用具体的视觉和音频细节进行描述，避免使用“cinematic（电影感）”或“beautiful（唯美）”等抽象词汇。 -
When using keyframes (I2VA / FL2VA / L2VA), clearly state how the first and/or last frame connects to the timeline.
使用关键帧模式（I2VA / FL2VA / L2VA）时，应明确说明首帧和/或尾帧如何与整个视频时间线衔接。