Guess the Sketch

a tangible, bite-sized way to inspire creativity

Create a game as a way to show creativity, with a bit of competition, and using Gen AI. Highlight AI/ML (Gen AI) realtime inference running on GKE on GPUs. This game was built as a demo to be shown at Next ‘24 and reused at future events and conferences.

Objective:

The game was built completely on Google Cloud using products that were highlighted throughout Next ’24. This helped land the message that anyone can build something end-to-end on Google Cloud.

Background:

The proposed game is the evolution of the classic game of telephone Pictionary, which is aptly described by Bard as follows:

Telephone Pictionary is a game that combines the verbal exercise of Telephone with the drawing game Pictionary.

Here's how to play Telephone Pictionary:

● Each player writes a phrase on the top sheet of paper.

● The next player draws a picture of their interpretation of the phrase.

● The next player captions the drawing.

● The game continues with players taking turns drawing and captioning until the paper is passed around and the message gets lost.

● When the game ends, you'll have a funny story written by you and your friends.

Telephone Pictionary is great for any group size, as long as they can be broken down into smaller groups of 5-10 people each.

In our rendition of the Next Gen AI Pictionary game, we used Generative AI to create images based on prompts written by the players.

The Game:

● Inspiring & tangible: A game is a tangible, bite-sized way to inspire creativity for using Gen AI within games. The proposed game is relatively simple to build and understand - developers can do this at home without large teams.

Impactful: a live playable game with a bit of competition is more engaging and has a longer lasting impact than a video or presentation.

Emphasizes the build on Google Cloud message: the game is built completely on Google Cloud using products that were highlighted throughout Next’24.

Why build a game rather than a video or a presentation?

Which products are highlighted?

● Game servers on the GKE Game Stack (Agones on GKE)

● Microservices on GKE Autopilot

● Inference on GKE on GPUs (stable diffusion)

● Inference on Vertex AI (Imagen)

● Google Gen AI model (Imagen)

UI Animation:

Wireframes:

There are 2 basic pages that are needed. Each page can have multiple states.

Requirements:

● P0: Should be compatible with mobile devices (tablet, smartphone, etc.), including portrait mode on mobile devices

● P1: On webpage, should display in landscape.

Homepage

Game Page 1: Initial Captions

Game Page 2: Guess the Image

Final Page: Scoreboard

A realtime dashboard showed the following metrics towards the end of the game:

a. Latency of image generation (prompt submission to image displayed)

b. Cost per session

c. Randomization of Gen AI models (or player chooses)

Loading Screen Animations:

Credits

Client: Google
PM: Ishan Sharma 
Production: jack morton worldwide

Scope

UI / UX
Motion Design
Previous
Previous

Google CES [UI Motion]

Next
Next

Shutterstock [UI Motion]