RaghuSasidhar/PromptPix — reverse-engineered prompt

Build me a simple web app called PromptPix where someone can use AI in two ways. First, they can upload an image and the app should describe it as a useful text prompt. Second, they can type a prompt and generate an image from it.

Make the experience clean and easy for a normal person. I want a home screen with clear sections for image to prompt and prompt to image, image previews, loading states, friendly error messages, and a way to copy the generated prompt. Use the Gemini API with an environment variable for the API key so I can run it locally without hardcoding secrets.

Please make it feel polished, not like a demo. Keep the layout responsive for desktop and mobile, and include simple instructions in the app so users know what to do. If you need the latest Gemini docs, look them up online.