Introduction

Machine learning applications like Stable Diffusion allow anyone to create images just by submitting a prompt consisting of a text description. This exercise gives you practice with one tool and invites you to ask what impact could this innovation have on creative people and careers.

👆 Sign up for a text-to-image app

1. Visit Gemini and sign in with your Google account (your school account works).

   ℹ️ As of fall 2026, Gemini's free tier generates roughly 20 images a day with its Nano Banana image model. Alternatives if you hit the limit: Leonardo.ai (about 150 free tokens/day) or Microsoft Designer.

   🎩 A frontier note: in 2024 this exercise required a dedicated image app, keyword-stacked prompts, and "negative prompts" listing what to avoid. Today's models accept plain sentences and follow them better — evidence of how fast the ground moves. The Medium/Subject/Details/Style/Composition vocabulary below still improves results, because it makes you more specific.

🗣 Prepare your prompt

In groups:

2. Decide who will answer each question (more than one person can answer the same question), and choose the following features of your image:
   • Medium
   • Subject
   • Details
   • Style
   • Composition

   Here are some examples of prompts that fit this template:

   • A digital painting of a snowy mountain landscape with vibrant pink and purple skies in the style of Impressionism with a focus on large, sweeping brush strokes and atmospheric lighting.
   • A charcoal sketch of an old man with a long beard and expressive eyes in the style of Renaissance portraiture with a close-up composition, capturing intricate facial details and emotion.
   • A pencil drawing of a futuristic cityscape with sleek, towering skyscrapers and flying vehicles in the style of cyberpunk art with a perspective from a high vantage point, emphasizing the city’s scale and neon glow.
   • A 3D render of an ancient Greek temple on a hill surrounded by olive trees in the style of classic architectural visualization with a low-angle composition to emphasize the grandeur and historical context of the structure.
   • A digital illustration of a mystical forest with glowing mushrooms and ethereal creatures in the style of fantasy art with a central composition that draws the eye to a shimmering river running through the scene.

👶 Submit naive prompt

3. Ask for an image in the chat (eg, “Please generate an image: …”) using at least one truncated version of your prompt that only includes Medium and Subject, eg:
   • A chalk pastel of a cozy cottage.

🧑‍🔬 Submit advanced prompt

4. In a new chat (so earlier context doesn't help it), submit at least one advanced version of your prompt that only includes Medium, Subject, Details, Style, and Composition, eg:
   • A chalk pastel of a cozy cottage by a lake under a starry night sky in the style of Impressionist landscapes with a wide-angled composition that captures the reflection of the stars on the still water.

📥 Post screenshots

5. As a group, submit two screenshots to #x-text-to-image. One should be your worst result and the other your best result, drawn from either the naive prompt or the advanced one.
6. Then vote on the best improvement from naive to advanced prompt (with 👍) and the worst (with 👎).