Landing Page
Host Lobby POV
Image Generation
User Prompt Creation
Round End
Leaderboard Screen

Prompt Perfect 🎨✨

The ultimate multiplayer challenge to master the art of prompting.

🌟 Project Inspiration

In an era where Generative AI is reshaping industries, prompt engineering has emerged as a fundamental bridge between human intent and machine execution. However, learning to "talk to AI" often feels like a solitary, trial-and-error process.

Prompt Perfect was born from the desire to gamify AI literacy. By turning the challenge of image recreation into a competitive social experience, we provide:

🚀 Impact & Usefulness

Real-World Relevance: As AI becomes ubiquitous, the ability to articulate precise descriptions is a high-value skill in creative, technical, and professional fields.
Potential Impact: It serves as an interactive playground for educators, hobbyists, and professionals to understand the nuances of diffusion models (how specific words change lighting, composition, and style).
User Value: Players aren't just playing a game; they are developing a mental model for how AI interprets language, all while having fun with friends.

🛠️ Tech Stack

Built with a high-concurrency architecture to ensure a seamless, real-time multiplayer experience.

Frontend: React 19 + Vite (JavaScript, HTML5, CSS3)
Backend: FastAPI (Python) with high-concurrency WebSocket support
AI Engine: Google Gemini (Imagen 3/4 for synthesis, 2.0 Flash Vision for semantic judging)
Real-time: WebSockets for instant game state synchronization
Design & Media: Figma (UI/UX) & Remotion (Programmatic video production)
Infrastructure: Vercel (Frontend & Serverless deployment)

✨ Features

Multiplayer Lobbies: Join friends via simple 6-character codes.
Live Game State: See who's ready, who's submitted, and who's currently "cooking" their prompts.
AI-Powered Scoring: Unlike simple pixel matching, our Gemini-based judge understands meaning. If the target is a "blue dragon," and you prompt "azure lizard," the judge recognizes the semantic closeness.
Dynamic Gallery: View every player's attempt side-by-side with the target image.
Global Leaderboard: Track scores across multiple rounds to crown the "Prompt Master."

🔄 Project Flow

Enter the Lobby: Players gather and prepare for the challenge.
Analyze the Target: A unique, AI-generated image is presented. Players have 60 seconds to "reverse-engineer" the prompt.
Submission: Players craft their best prompt and submit.
The Reveal & Judging: The server generates images for all players simultaneously. The Gemini Vision Judge compares each attempt to the target.
Victory: The final leaderboard shows who best understood the AI's "brain."