Skip to content

Knowledge that lights the world

Creating Images With AI: Tools, Writing a Good Image Prompt, Copyright and Ethical Questions

Producing photorealistic images from a few sentences is now within everyone’s reach. How image generators work, prompt tips for good results, commercial use, copyright and the risks of fake images.

Creating Images With AI: Tools, Writing a Good Image Prompt, Copyright and Ethical Questions

AI tools that generate images from text spread rapidly from 2022 onwards. Most work with a method called “diffusion”: during training the model learns how noise is added to images step by step and then how to remove it. When generating, it starts from random noise and “cleans” the image step by step to match your description.

Main types of tools

  • Chat assistants’ image features: Assistants such as ChatGPT, Gemini and Copilot let you create and edit images within a chat, which is handy for quick experiments.
  • Dedicated image tools: Tools such as Midjourney stand out for artistic style and detail control.
  • Tools built into design software: Tools such as Adobe Firefly offer edits like extending existing photos or removing objects.
  • Open models: Models such as Stable Diffusion can run on your own computer but need more technical knowledge.

How to write a good image prompt

A vague prompt (“a nice landscape”) gives random results. For better results, specify:

  • Subject: “A small stone house with a red roof”
  • Setting: “On a misty mountainside in autumn”
  • Style: “Watercolour painting” or “realistic photograph”
  • Light and colour: “Sunset light, warm tones”
  • Composition: “Wide angle, the house in the lower right corner”

If you don’t like the result, change one element at a time rather than everything at once.

Commercial use and copyright

Commercial rights to generated images depend on the tool’s terms of use; some tools allow commercial use on paid plans. Copyright is legally contested. The US Copyright Office, for example, has stated that content generated entirely by AI without human contribution cannot be protected by copyright. Some artists and image agencies have also sued, arguing their works were used as training data without permission. For commercial projects it is important to read the tool’s licence terms and seek legal advice if needed.

Ethical rules

  • Don’t create realistic images of real people without their consent.
  • Don’t share fake images that look like news or documents; state that they were made with AI.
  • Develop your own style rather than copying living artists exactly.
  • Don’t use brand logos or protected characters without permission.

How do image-generating models work?

Most of today's image generation tools are based on a method called a "diffusion model". During training, the model sees millions of image-text pairs; random noise is added to the images step by step, and the model is asked to clean that noise away again. At generation time the process runs in reverse: the model starts from pure noise, uses the textual meaning of your prompt as a guide, removes the noise over dozens of steps and finally produces a new image.

This approach became widespread with the paper "Denoising Diffusion Probabilistic Models", published in 2020. It is also why the same prompt gives a different result each time: the starting noise is different for every generation.

How can AI images be recognised? Labelling and content credentials

Realistic AI images can no longer always be told apart with the naked eye. This is why technical labelling methods are coming to the fore:

  • C2PA content credentials: An open standard developed jointly by technology and media organisations adds a signed record to the file showing how an image was created and edited.
  • Legal obligation: Article 50 of the EU Artificial Intelligence Act requires AI-generated content to be marked in a machine-readable way and realistic fake content (deepfakes) to be clearly disclosed.
  • Visual check: Inconsistencies in hands, text, reflections and background details can still give clues, but they should not be treated as proof on their own.

Adding an "AI-generated" note when sharing images you create both protects trust and keeps you in line with the rules of many platforms.

AI tools’ features, plans and prices change often. Always check the tool’s official page for current information.

Frequently asked questions

How can you tell an AI-generated image?

Inconsistencies in hands, text and background details can be clues, but these signs are fading as tools improve. Some companies add invisible watermarks or content credentials to images, and reverse image search can help find the source.

Sources and further reading

  1. U.S. Copyright Office, “Copyright and Artificial Intelligence, Part 2: Copyrightability” (January 2025).
  2. Robin Rombach et al., “High-Resolution Image Synthesis with Latent Diffusion Models”, CVPR, 2022.
  3. Coalition for Content Provenance and Authenticity (C2PA), content credentials standard.
  4. Ho, J., Jain, A., Abbeel, P. (2020). Denoising Diffusion Probabilistic Models. NeurIPS 2020
  5. C2PA – Content Credentials Technical Specification
  6. EUR-Lex – Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 50
Share
Follow YulaHubGet the day's most important stories first.
RSS

Leave a comment

Your email address will not be published. Required fields are marked *