AI image generation also gave rise to the "prompt engineering" concept. Some users keep their combinations under lock and key, while others share them openly, but the point is that there is always something new to learn — and that is how we arrive at Say What You See, an experiment created by artist Jack Wild for the Google Arts & Culture Lab. The rules are simple: look at the image, create a prompt, follow the recommendations, and level up!
One of the most debated aspects online about AI image generation is the real influence of the prompt, and as you can imagine, the number of opinions is gigantic. Some believe its impact is minimal, while others stand on the opposite side and consider it essential. The type of model, its prior optimization, the seed number, the intensity and other secondary parameters can make a prompt obeyed to the letter or ignored almost completely. However, today we have a different proposal: turning prompt creation into a game.
How to create better prompts with Say What You See
The process is very simple: Say What You See presents a generated image. Your mission, if you decide to accept it, is to describe (in English) the content of the image with a prompt. There are three images per level, and each one raises the "similarity threshold", but the game always invites us to explore specific details and other modifiers. At the highest levels, Say What You See shows no mercy (for example, level 7 wants 90 percent precision), but the experiment lets us advance even if we make mistakes.
Say What You See is a creation of Jack Wild, resident artist at the Google Arts & Culture Lab. It has been online for several months (available data suggests it debuted last December), but it seems to have gone somewhat viral in recent hours. Link below!
Official site: Click here
