How do I generate images?

Dartmouth Chat can be used to generate images from prompts. Models are trained on massive datasets of images with the goal of learning patterns used by humans to understand their surroundings.

Within Dartmouth chat click on the “Image” icon under the “Send a Message” area to specify that you want to generate an image.

Similar to text prompts, the specificity of your prompt will define the quality and accuracy of the output. Some things to consider when creating a successful prompt:

• Set the Scene. Provide specific information about the scene - e.g., “A bright and clear daytime photo of Dartmouth College’s Berry-Baker Library lawn, with the iconic red-brick building and white clock tower in the background.”

• Define the Style. Identify the style to be used - e.g., “A high-quality photograph.”

• Provide Precise Detals. Provide clear instructions about what’s happening in the image - e.g. “In the foreground on the grass, there is an outdoor classroom setup with a U-shaped arrangement of wooden benches or modern outdoor chairs. Around 12 students of diverse backgrounds are seated, some with laptops and notebooks, actively participating in a lesson. One instructor stands at a portable whiteboard, gesturing as they teach. The students are engaged, with some raising their hands. There are backpacks, water bottles, and textbooks on the ground.”

• Provide Directions about Mood or Emotion. Explain the emotional tone are you aiming for - e.g., “The overall mood is vibrant, dynamic, and scholarly.”

• Provide Composition or Color Palette Instruction. Define limitations or expectations clearly - e.g., “The resulting square image should utilize bright, natural colors.”