AI GUIDEWhat you can make
You do not need to know illustration software to explore how an idea might look. An image generator can propose a poster concept, a scene for a story or a different arrangement of objects. An editing model can use an existing image as a starting point and attempt a requested change.
These are useful for exploring options before spending time on a finished design. They are also different from image understanding: a model that describes a photograph is not necessarily able to produce or edit one. Check the chosen service's image features rather than assuming every chat has them.
Our exercise is a fictional poster for a neighbourhood reading afternoon. Use invented scenes and no real person's photograph. The goal is to communicate a warm, quiet gathering, not to manufacture a photograph of an event that actually happened.
Write a visual brief
A useful brief explains the purpose, subject, composition, mood and constraints. Try: “Create a portrait-format illustration for a fictional neighbourhood reading afternoon. A small table with books and two empty chairs beside a sunny window. Warm cream and muted blue colours. Leave a clear area at the top for a title added later. No logos, no readable text, no people. The image should feel welcoming rather than luxurious.”
This is an original prompt, not a tested model output. Asking for empty title space is a practical design choice: it lets you set the final wording accurately in an editor. It is not a guarantee that every generated composition will leave enough space.
Google's image guidance supports describing a scene and using reference images for generation or editing. Choose a tool that explicitly offers the operation you need. Current model names, quotas and prices belong in the linked product documentation, not in a permanent promise inside the lesson.
Direct a small change
Suppose the first concept has a cluttered table. If the tool supports editing, use the selected image and ask: “Remove the loose papers. Keep the books, chairs, window, colours and overall layout unchanged.” This describes both what to change and what to preserve.
Inspect the whole result. A local edit can change other objects, lighting or proportions. If the service provides separate generation and editing modes, make sure you are using the intended one. Repeating the original text prompt may generate a new composition instead of editing your chosen version.
For a client product image, visual similarity is not proof that colour, dimensions or printed labels are accurate. Keep an original reference and check important details individually. For our fictional scene, the relevant check is whether the composition still serves the poster.
Judge and finish
Choose a concept using the brief: can the title fit, is the intended mood clear, and are the objects understandable at the final display size? A surprising image is not necessarily a useful one.
Add the event title, date and location in a layout editor after checking their accuracy. If sharing the exercise, label it a fictional concept. Do not describe an AI illustration as a photograph of a real gathering.
Exercise: make two briefs with the same content but different moods, then choose which better invites a quiet reading audience. Explain the choice with composition and purpose, not “the AI said it was best”. If you have no image tool, sketch the composition and test the brief with another person. You can practise the design decision before paying for generation.
Sources
Read the original sources to explore the topic and check specific claims.
Google: image generation and editing guidance ↗