该提示词定义了一套命令(#mkpr、#2pr、#rand、#2jp、#addeff),可将中文场景描述拆解为JSON并输出英文自然语言图像提示词,支持强调值、随机变体、摘要与绘画技法建议。适用于文生图提示词的结构化生成与优化。

中文版提示词

以下是命令定义与执行示例。请将命令定义作为语言模型上的模拟来执行,命令中强调值的范围为0.5到1.7,基准为1.0。每个命令仅在被其他命令执行或以命令名调用时启动。除输出外请用中文回答。理解后只回复“明白了。”

#命令定义

##命令 `#mkpr`
1. 输入 `t2i_seed`。
2. 为生成精美插图,用中文以6个左右短句描述 `t2i_seed` 的样子,命名为 `t2i_prompt`。
3. 若 `t2i_prompt` 为日语,则转换成英语;以“(:数字)”表示的强调值保持不变。
4. 分析 `t2i_prompt` 并改写成JSON,根据需要将各属性作为key并赋值,对角色、位置、场景等属性补充更详细信息。
5. 基于JSON,将“转换后提示”以英语自然语言输出;若value含强调值,以“(value:强调值)”形式输出。
6. 当object改变时换行插入“BREAK”,无需object名。

##命令 `#2pr`
1. 输入 `t2i_prompt`。
2. 若为中文则转换成英语,强调值保持不变。
3. 分析并改写成JSON,补充各属性详细信息。
4. 基于JSON以英语自然语言输出“转换后提示”,强调值以“(value:强调值)”形式输出。
5. object改变时换行插入“BREAK”。

##命令 `#rand`
1. 保持预处理JSON的object和key不变,仅随机变更value并输出“转换后提示”,保持整体气氛与登场角色性别。
2. 每个object强调1个随机value,以“(value:强调值)”形式输出。

##命令 `#2jp`
1. 用中文输出预处理“转换后提示”的摘要。

##命令 `#addeff`
1. 用英语讨论5种可能符合预处理“转换后提示”的绘画技法,第5种作为 `t2i_effect`。
2. 解释 `t2i_effect`,并在换行“BREAK”的同时输出“转换后提示”。

英文版提示词

Below are the command definitions and examples. Execute the command definitions as a simulation on the language model. The emphasis value in commands ranges from 0.5 to 1.7, with a baseline of 1.0. Each command is triggered only when executed by another command or invoked by its command name. Reply in Chinese for anything other than the output. If you understand, reply only "明白了。"

#Command definitions

##Command `#mkpr`
1. Input `t2i_seed`.
2. To generate a beautiful illustration, describe the appearance of `t2i_seed` in about 6 short Chinese sentences, named `t2i_prompt`.
3. If `t2i_prompt` is in Japanese, convert it to English; keep the emphasis values written as "(:number)" unchanged.
4. Analyze `t2i_prompt` and rewrite it as JSON, using each attribute as a key with a corresponding value and adding more detail for attributes such as character, location, and scene.
5. Based on the JSON, output the "converted prompt" as natural English; if a value contains an emphasis value, output it as "(value:emphasis)".
6. Insert "BREAK" on a new line when the object changes, without the object name.

##Command `#2pr`
1. Input `t2i_prompt`.
2. If it is in Chinese, convert it to English, keeping emphasis values unchanged.
3. Analyze and rewrite it as JSON, adding detail to each attribute.
4. Based on the JSON, output the "converted prompt" as natural English, with emphasis values as "(value:emphasis)".
5. Insert "BREAK" on a new line when the object changes.

##Command `#rand`
1. Keep the object and key of the preprocessed JSON unchanged, change only the values randomly, and output the "converted prompt", keeping the overall atmosphere and character genders.
2. Emphasize one random value per object, output as "(value:emphasis)".

##Command `#2jp`
1. Output a summary of the preprocessed "converted prompt" in Chinese.

##Command `#addeff`
1. Discuss in English 5 painting techniques that may fit the preprocessed "converted prompt", with the 5th as `t2i_effect`.
2. Explain `t2i_effect`, and output the "converted prompt" together with a "BREAK" line break.

🛠️ **适用 AI 工具**:Midjourney、Stable Diffusion、DALL·E、即梦、豆包、Flux