Say What You See was a delightful mini-game introduced by Google Arts & Culture in November. It educates users on the “art of picture prompting” and serves as effective training.
The game presents an AI-generated image on the left, challenging players to write a concise caption of up to 120 characters for an accurate match. After the 30-second round, players can explore artworks from the Arts & Culture collection. Google has not disclosed the specific model used for this game. Imagen 2, launched in December, boasts “markedly enhanced image quality,” characterized by a “photorealistic, high-resolution, and visually appealing” aesthetic.
Players are allowed three attempts to pass a visual similarity threshold in Level 1. Google offers guidance on refining prompts, suggesting details about the medium, subject, context, art period, materials, shapes, and textures. Level 2 intensifies the challenge, requiring players to fulfill 50% of the criteria.
The game is accessible via the Arts & Culture app, with availability expanding gradually to more countries. Additionally, participants in Google Search Labs can utilize the Search Generative Experience (SGE) to generate image prompts post-game.