For better AI images in Microsoft Copilot, use prompts that are specific, descriptive, and iterative.
Recommended techniques for beginners:
- Be descriptive instead of generic
- Weak:
Create an image of a lion - Better:
Create a close-up photo of a female lion in a vast grassland with multiple shades of green and brown
- Weak:
- Add visual details
Include:
- adjectives
- location or setting
- artistic style
- realism level
-
digital art -
photorealistic -
realistic -
bright colors -
film stock
- State clearly whether the goal is a stock image or a generated image In PowerPoint, asking Copilot to “add a stock image” adds a stock image, while asking it to “generate an image” uses DALL·E 3 to create a new image.
- Give Copilot context about the result wanted
Strong prompts usually include:
- the goal
- relevant context
- expectations for the output
- the source or subject to use
- Use action words
Prompts work better when they clearly say what to do, such as
create,filter, orsummarize. For images, start with direct verbs likeCreate an image of.... - Use correct and relevant terminology Use clear terms for the subject and avoid vague wording, misspellings, abbreviations, or unnecessary symbols.
- Iterate and refine If the first result is not right, adjust the prompt and try again. Refining details often improves the output.
- Use English for best results Copilot features work best with English prompts.
Beginner prompt formula:
Create a [style] image of [subject] in [setting], with [key details], [lighting/color/mood], and [visual style].
Example:
Create a photorealistic image of an astronaut in a glowing plasma suit, putting on a helmet, with bright colors and a cinematic film-stock look.
If a brand-related image is needed in Word, include the word brand in the prompt.
References: