Puzzle AI-powered
Say What You See
Describe an AI-generated image, then refine your prompt to produce a closer visual match.
By Jack Wild with Google Arts & Culture Lab
Opens the creator’s website in a new tab.
Listing reviewed
About Say What You See
Say What You See turns image prompting into a visual puzzle. Study a reference image, describe what you notice, and use that description to generate a new image. The challenge is to make the result sufficiently similar to pass the current level. Google's instructions allow three attempts per image and provide tips as the difficulty increases.
The AI is part of the activity itself: your text prompts another image, rather than simply selecting a prewritten answer. This makes the experiment a way to practice observing subjects, composition, and descriptive details. It was created by artist Jack Wild with Google Arts & Culture Lab. The play link opens Google's official experiment. Our launch check reached the first level and its prompt field in Chrome; we did not submit a description or test image generation.
How to play
- Text description
- Describe the reference image to generate a matching image
- Prompt revision
- Refine your description within the available attempts
Where AI comes in
Google's official description says player descriptions prompt generated images, which must meet a visual similarity threshold to pass a level.
Read the sourceBefore you play
Official experiment and announcement reviewed. On 2026-10-11, a Chrome launch check reached Level 1, its 50% similarity target, and the reference-image prompt screen. The browser was already signed in. No prompt was submitted; generation, anonymous access, and availability in other regions remain unverified.
