Creating Images with ChatGPT

ChatGPT can generate original images and quickly modify or edit them using only plain natural language prompts. The key is not to write at length, but to clearly convey the purpose, subject, mood, scene, and style of the image within 1 to 3 sentences.

A good prompt specifically states what to show, how it should look, and what it needs to achieve. If elements like composition, lighting, materials, or size matter, it's far more reliable to specify them directly, and when there's something you don't want, it's effective to clearly state constraints such as "no logo" or "no additional text."

When editing an image, the best approach is to iterate in small steps. First get the core idea right, then adjust one element at a time—such as brightness, color tone, simplifying the background, or changing the style—so the results stay more stable.

It's also possible to upload multiple images together as reference material. In this case, refer to each image in order and use spatial expressions like left, right, foreground, and background to describe the relationships between them.

For images that include text, instructions need to be even stricter.

• Wrap the text in quotation marks or write it in uppercase
• Specify the font, size, color, and position
• Keep the wording short
• Spell out brand names or unfamiliar words letter by letter

For text-heavy compositions like infographics, posters, diagrams, and timelines, it's better to emphasize sharp text rendering and, if needed, finish the piece in a design tool. When dealing with real people's faces, portrait rights and consent should be considered, and rather than imitating a specific brand or work, it's safer to request a generic, original design. Generated images don't necessarily require separate credit, but you should follow your organization's policies and OpenAI usage policies.