Image generated with Nightcafe. Prompt: "A modern, dark texture suitable for Web headers that reminds us of cameras and photography."
Details matter to generate the best possible experimental stimuli.
This is a Web version of the Companion Appendix for RAISE: A New Method to Develop Experimental Stimuli for Advertising Research with Image Generative Artificial Intelligence, developed by César Zamudio, Jamie Grigsby, and Meg Michelsen, and published in the Journal of Advertising. Questions? Feel free to contact us!
When constructing Stimulus A, we recommend establishing a number of details to be kept consistent across stimuli.
Aspect ratio
Aspect ratio, which must be decided before constructing the prompt, is differently specified across tools. Some tools (e.g., Dreamstudio) include an aspect ratio selection tool, while others (e.g., Midjourney) instead determine aspect ratio in the text prompt itself (e.g., “a busy park with a beautiful fountain ar -- 1:1” for a square aspect ratio, where “ar –" is a prompt element specific to this platform). Decisions concerning aspect ratio might depend on the devices that experiment respondents are likely to use. Additionally, wider aspect ratios (e.g., 16:9) might be better suited for panoramic stimuli, while narrower, elongated ratios (e.g., 2:3) might be better suited to generate full body pictures of people. or to better emulate a real print advertisement.
Artistic style
Artistic style gives the image generative AI its “look and feel”, and researchers should therefore decide whether artistic style must be specified (for instance, if desiring a realistic depiction of a person, a “photographic” style might be used). Artistic style can be induced using two methods which can be used separately or simultaneously. One method is to request a style from a pre-existing list available within the AI tool itself – for example, DreamStudio offers preset styles such as “Photographic” or “3D model”, while Dreamlike features artistic styles associated with different image generative AI tool models, such as “Neurogen” or “Dreamlike Photorreal” – in this case, visual examples provided by each tool might dictate if any pre-determined styles might be used.
A second method is to include the artistic style as part of the text prompt used as input to the image generative AI tool, requesting a specific style, e.g. “Create a picture of [character] in the style of [artist]” (Wang et al., 2023). Below, we present how these two methods can be used and resulting example images. Note that RAISE is not used in this case, as this example only illustrates artistic styles.
A second method is to include the artistic style as part of the text prompt used as input to the image generative AI tool, requesting a specific style, e.g. “Create a picture of [character] in the style of [artist]” (Wang et al., 2023). Below, we present how these two methods can be used and resulting example images. Note that RAISE is not used in this case, as this example only illustrates artistic styles.
Camera details
When constructing the initial text prompt to be used with RAISE, general terms related to photographic quality, such as “high-quality photo”, or “8k”, might result in output images of different quality, and thus experimentation is recommended. However, beyond these simple terms, researchers familiar with more formal terms associated with photography (e.g., f-number aperture, fisheye lens, bokeh effect, macro, and so forth), might benefit from explicitly including such details in image generative AI text prompts as well.
For instance, researchers might want to produce a stimulus in which the foreground is focused on, with an aesthetically pleasing, out of focus background featuring motes of light –a bokeh effect. Below, we provide examples of varying image characteristics resulting from using photography terms. Note that RAISE is also not used in this case, as this example only illustrates camera details.
For instance, researchers might want to produce a stimulus in which the foreground is focused on, with an aesthetically pleasing, out of focus background featuring motes of light –a bokeh effect. Below, we provide examples of varying image characteristics resulting from using photography terms. Note that RAISE is also not used in this case, as this example only illustrates camera details.
Ready to learn more?
Feel free to browse our gallery of examples with full tutorials, and the technical details section to learn the finer points of generating visual experimental stimuli using the RAISE methodology.