Website at copilot.microsoft.com/ (use of FREE version).
Image creation here is purely by prompting as shown. Note: Copilot is Microsoft's customized application of OpenAI's technology, though not an identical model.
0. The initial screen of Microsoft Copilot.
1. Prompt: "Please give me an image of a traditional teacher at a blackboard"
2. Prompt: "Give it a cartoon image style"
3. Prompt: "Have the word 'Dunce' written on the blackboard"
4. Prompt: "Add a cat sitting on the floor with the teacher looking at it"
5. Prompt: "Please give me a different version of this". This produced a version that was very similar except for subtle things like eye direction and skirt length.
6. Prompt: "Give me a version where we see more of the two sides of the room".
7. Prompt: "Add some bored looking cats on desks at the sides of the room". Copilot refused to generate this image, even when rephrased, saying things like: 'I can't create that image due to a policy restriction'. It repeatedly produced 90% of the image which was perfectly nice and normal with multiple bored looking cats before removing it with the above message.