The most frustrating part of creating a text-and-image report is the images: searched images are authentic but inconsistent in style, while generated images are consistent but require good descriptions. You don't have to choose one—MuchSci Office Assistant supports both routes: search and save images from the public image library, and have AI generate images based on descriptions (with specified dimensions and aspect ratios). Choose the route based on the content type, then combine everything into the same PDF.
Route 1: Image Search—for Documentary Content
Tell MuchSci Office Assistant: "Search for a few images of XX city at night and save them for use." Images obtained from the public image library have an authenticity that generated images cannot provide: event photos, real product shots, cityscapes, news scenes, and documentary portraits should all take this route—audiences have a natural trust in "this was actually photographed."
Three key points:
l Keep horizontal and vertical orientations as consistent as possible so the layout doesn't look crooked when mixed;
l Choose similar color tones for images of the same type, which cuts the inconsistent-style problem in half;
l Record the source and note it in the report—this is an obligation of the image-search route, not optional.
Route 2: Image Generation—for Conceptual Content
Tell it: "Generate a cover image for the theme XX in 16:9." The advantage of AI generation is that it follows your description completely: style, color tone, and ratio can all be specified. Abstract concepts that can't be drawn, such as "growth," "connection," "efficiency," and "collaboration," can't be found appropriately through image search, but can be generated on demand with a prompt.
Three key points:
l State three things clearly in the description: theme + style + ratio; missing any one increases randomness;
l For generated images in the same report, use a similar description template so the whole set looks like it was drawn by one person;
l The cover key visual is worth generating two or three options to choose from; small in-body images can be done in one pass.
Figure 1 | Choosing a route by content type: documentary content uses image search; conceptual content uses generated images.
Convergence: Arrange Into a PDF
Once the images from both routes are ready, hand them over together with the text: "Generate a text-and-image report PDF." PDF creation, editing, page operations, and Chinese typesetting are its finishing capabilities—the cover, text-and-image paragraphs, and source citations are completed in one go.
There is a useful rhythm on the page: one image + one explanation, arranged alternately.Images are not decoration; every image should be "used" by the body text—the question a reader has after viewing an image is answered by the next sentence, and the report holds together.
Figure 2 | Finished structure: cover, text-and-image paragraphs, source citations.
Mixed-Use Example in One Report
Take "A Research Report on a City's Nighttime Economy" as an example: the cover key visual uses a generated image (16:9, neon color tone); Chapter 1's current situation uses searched real city scenes; Chapter 2's data section generates one conceptual supporting image; Chapter 3's cases all use searched real photos; sources are noted consistently at the end.Forty percent generated images and sixty percent searched images keep both authenticity and consistency—the ratio is not fixed; adjust it according to the nature of the content.