Why it matters
Image AIs often get the number of objects wrong. We asked for ten doors and got twelve. Here is a quick way to catch it.
The 5-step check
0 of 5 doneTap a step when you have done it. Your ticks stay on this device.
the AI makes it. You count it.
Sources: Guo et al., "Text-to-Image Diffusion Models Cannot Count, and Prompt Refinement Cannot Help" (arXiv 2503.06884); Kang et al., "Counting Guidance for High Fidelity Text-to-Image Synthesis" (arXiv 2306.17567). The door examples are our own experience. Step 1 is our own suggestion. Steps 2, 3 and 5 are general good practice.




