This is a submission for the Cloudflare AI Challenge.
What I Built
I build a guess the prompt app using AI generated images. Playing Frenzy is a breeze. Upon starting the game, player is presented with a randomly generated image and a set of title options. Using their intuition and deductive skills, player must select the correct title that best describes the prompt to generate the image. With each correct guess, player earn points keeping them engaged and motivated to continue playing.
Demo
My Code
Github Repo Link: https://github.com/raibove/frenzy
The code for workers is in the came repository under workers folder.
In workers I have added two api
-
/optionswhich take query as params and generate 4 options for given query. -
/imagewhich take title and generate image for given prompt.
Journey
I always wanted to try creating a app using AI models and cloudflare challenge was a perfect opportunity to do so. Before starting the project I explored how to use workers and tried different models in the playground. I then finalised a topic and started working on it.
The concept behind Frenzy is simple yet innovative:
I first select a random topic like "fantasy creatures", "celestial bodies", "emotions", "time", "weather" and generate 4 options using llama model then I select any one option and generate a image for it using Stable Diffusion.
Player then has to guess which prompt was used to generate the image.
I have plan to expand frenzy to a multiplayer game and add options for playing in different languages and using object detection to generate AI based hints.
Multiple Models and/or Triple Task Types
I used Multiple Models per task for generating the quiz I am using stable diffusion and llama.
Models used
You can try playing frenzy in this link. Let me know in the comments if you face any issues or like my idea.
You can connect with me on twitter - https://twitter.com/shweta_kale1
Top comments (0)