Creating images with ChatGPT

ChatGPT 可以通过普通语言提示生成原创图像。你可以快速迭代——请求变体、调整构图或尺寸、探索新的视觉方向——并在几分钟内产出可直接投入使用的素材。

这使得探索概念、用视觉方式传达想法，以及为不同受众、格式或渠道调整现有素材都变得更容易。

撰写有效的图像提示

一个好的图像提示不需要很长。在大多数情况下，1–3 句清晰的话就足够了。目标是让 ChatGPT 明白这张图像是什么、应有的氛围，以及它需要达成的目的。

实践中，这意味着用少量关键细节来给提示打底：图像的用途、主要主体、发生了什么、发生地点，以及期望的视觉风格。如果构图、光线或特定约束很重要，也要一并写明。

清晰比华丽措辞更有效——尤其是像布局、纹理、材质或光线这样的细节。例如，“从左侧窗户来的柔和自然光”通常比“漂亮的灯光”更可靠。

当某些元素必须保持不变时，约束条件格外有用。如果你不希望出现额外文字、标识或视觉改动，就直接说明。编辑现有图像时，要明确指出应改动的内容与应保持不变的部分。像“只更改 X，其他全部保持不变”这样的提示，通常是指导精确编辑最清楚的方式。

试试看

创建一幅简洁但精致的社论插图：有人在桌前学习新的 AI 技能。画面应包含笔记本电脑、笔记本、书籍、便签，以及几个微妙的进展标识，例如已勾选的复选框、高亮的段落，或别着的有条理的计划。人物应显得专注且投入，整体场景给人平静、专注、现实的感觉。使用干净、极简的背景和现代的数字插画风格，既亲切又中性。避免出现标识或品牌元素，也不要用科幻或过于抽象的意象。

最佳实践以提升效果

通常通过小而有针对性的修改可以最有效地改进图像。先把核心想法确定好，然后一次调整一个元素。直接、具体的反馈比笼统的反应更容易执行；重复最重要的细节也有助于在反复打磨时防止图像偏离。

一些有用且可操作的调整示例：

• “调亮一点”
• “弱化颜色”
• “简化背景”
• “保持相同构图，但把风格改得更现代 / 更柔和 / 更俏皮”

按步骤逐项修改有助于在细化过程中保持一致性。你也可以指定图像的某些区域并给出有针对性的指令。

高级指南

使用多张上传图片

你可以上传多张图片来引导生成或编辑，但通常较少的图片更易于管理。按顺序引用每张图片并说明它们之间的关系。

例如：“ Image 1 是我的桌面设置照片。 Image 2 是风格参考。把 Image 2 的干净、极简插画风格应用到 Image 1，同时保持相同的布局和物件。”

在合成元素时，使用清晰的空间语言——如左边、右边、前景、背景——来描述各元素之间的关系。

图像中的文字（标题、标签、主标题）

对文字的指令要非常具体：

• 把文字放在引号里或使用全大写
• 指定字体风格、字号、颜色、位置
• 文字内容要简短
• 对于品牌名或不常见词汇，逐字母拼出（例如 “ S-T-R-I-P-E ”）

例如：在顶部居中以粗体无衬线、白色、72pt 添加标题 “ WEEKLY PLAN ”。不要添加其他文字。

信息图与密集布局

信息图适合用于说明性内容、海报、有标注的图解、时间线和“视觉百科”类素材。对于文字较多或布局密集的图像，要强调“文字清晰渲染”，并考虑通过设计工具做后期润色。

其他注意事项

• 对肖像保持谨慎：生成真实人物（包括你自己或他人）的图像时，使用参考照片以保证准确性，并确保你有权使用他们的肖像。
• 泛化优于模仿：如果不确定，优先要求“通用”或“可自有化”的设计，而不是模仿某个特定品牌、产品或艺术作品。
• 署名可选：使用生成图像时不要求一定要标注 OpenAI，但如果注明有助于说明素材来源，你也可以这么做。
• 遵守政策：所有图像使用应遵循你所在机构的规范以及 OpenAI 的使用政策。

通过 OpenAI Academy 持续学习

查看更多指南和资源，帮助你构建实用的 AI 技能。参见 OpenAI Academy 的全部主题。

ChatGPT can generate original images from plain-language prompts. You can iterate quickly—request variations, adjust composition or size, or explore new visual directions—and produce production-ready assets in minutes.

This makes it easier to explore concepts, communicate ideas visually, and adapt existing assets for different audiences, formats, or channels.

Writing effective image prompts

A good image prompt does not need to be long. In most cases, 1–3 clear sentences are enough. The goal is to help ChatGPT understand what the image is, how it should feel, and what it needs to accomplish.

In practice, this means grounding the prompt in a few key details: the purpose of the image, the main subject, what is happening, where it takes place, and the desired visual style. If framing, lighting, or specific constraints matter, include those too.

Clarity is more effective than clever phrasing—especially for details like layout, texture, materials, or light. For example, “soft natural light from a window on the left” will usually be more reliable than something vague like “beautiful lighting.”

Constraints are especially useful when something needs to stay fixed. If you do not want extra text, logos, or visual changes, state that directly. When editing an existing image, be explicit about what should change and what should stay the same. A prompt like “Change only X. Keep everything else exactly the same” is often the clearest way to guide a precise edit.

Best practices to improve results

The best way to improve an image is usually through small, targeted revisions. Start by getting the core idea right, then adjust one element at a time. Direct, specific feedback is easier to follow than broad reactions, and repeating the most important details can help prevent the image from drifting as you refine it.

Examples of useful, actionable adjustments:

• “Make it brighter,” “tone down the colors,” “simplify the background”
• “Keep the same composition, but make the style more modern / softer / more playful”

Step-by-step revisions help maintain consistency as you refine the image. You can also edit specific areas and provide targeted instructions.

Advanced guidance

You can upload multiple images to guide generation or editing, but a small set is usually easier to manage than a large one. Refer to each image by order and explain how they relate to each other.

For example: “Image 1 is a photo of my desk setup. Image 2 is a style reference. Apply image 2’s clean, minimal illustration style to image 1, while keeping the same layout and objects.”

When combining elements, use clear spatial language—such as left, right, foreground, and background—to describe relationships.

Text works best when instructions are very specific:

• Put text in quotes or ALL CAPS
• Specify font style, size, color, placement
• Keep text short
• For brand names/uncommon words, spell out letter-by-letter (e.g., “S-T-R-I-P-E”)

For example: "Add the headline “WEEKLY PLAN” in bold sans-serif, white, centered at the top, 72pt. No other text."

Infographics are useful for explainers, posters, labeled diagrams, timelines, and “visual wiki” assets. For dense layouts or heavy in-image text, emphasize “sharp text rendering,” and consider polishing in design tools if needed.

Additional considerations

• Be mindful with likenesses: If generating images of real people (including yourself or others), use a reference photo for accuracy and make sure you have permission to use their likeness.
• Generic over specific: When in doubt, request “generic” or “ownable” versions of a design rather than imitating a specific brand, product, or artwork.
• Attribution optional: You aren’t required to credit OpenAI when using generated images, but you may do so if it helps clarify how the asset was created.
• Policy compliance: All image use should follow your organization’s guidelines and OpenAI’s usage policies⁠.

Continue learning with OpenAI Academy

Generated by RSStT. The copyright belongs to the original author.

Source