AI tools can be used to generate images in the same way that natural language commands can be used to generate text in ChatGPT. Image generators do not only work based on the material used to teach the system; it is equally important to be able to define the prompt in such a way that the desired result is produced.
How to do? The AI is guided by a prompt, which is used to generate the answer. The prompt should describe in more detail not only the subject matter but also the style in which the image will be created. Terms such as illustration, photorealistic, high definition, pencil drawing, water colour and oil painting, graphic art, 20's style, etc. can be used as a definition. See here for further guidance on how to make good prompts for generating images.
It takes a while to make the images. Instead of a single image, the user is given nine thumbnail images. The machine also generates different images each time, even if the feed is exactly the same.
Use of the image in the text: the image must state that it was generated by an AI tool, which tool it is and which prompt was used.
The key to generating images is to be able to make good and descriptive prompts. Phraser is an app that helps you write prompts adapted to your purposes and to different neural networks (DALL-E 2, Midjourney, Craiyon, etc.).
Phraser breaks down the process of creating a good prompt into small steps - choosing the type of content, style, emotion, era, quality and so on.
Figure 1: Image generated by Open AI:n DALL-E with the command Cat doing information retrieval graphic art.
Laurea-kirjasto | Saavutettavuusseloste | Laurea Library | Accessibility statement