nithishnaidu918/stable_diffusion ? reverse-engineered prompt

Build me a simple multimodal app for BMW images.

I want two things in one place, first a text to image generator using Stable Diffusion, and second an image to text feature that can look at a BMW photo and give me a detailed description of what it sees. Make it easy to run locally, with a clean interface for entering a prompt, uploading an image, and seeing the generated result or description right away.

It should use the GPU when available, but still work on CPU if not. If the models need to download the first time, that’s fine, just handle it smoothly and show clear status while it loads. Keep the experience straightforward and polished, and save generated outputs somewhere useful so I can find them again.

If you need to check current model setup details or best practices online, go ahead and look them up.