.AIGX

🧠 Inspiration

I was inspired by how WebP made the web faster by compressing images better than JPEG or PNG. But with the rise of AI-generated content, I realized there's no format designed specifically for AI media. Traditional formats waste storage, energy, and can’t store the prompt, model, or allow easy reuse. That’s why I created .AIGX — a new image format for AI-generated images that’s lightweight, traceable, and regenerable.

🚀 What it does

.AIGX is a custom file format (like .zip) that includes:

• The prompt, seed, and model version
• The latent vector (instead of raw pixels)
• A thumbnail preview
• Optional masks, metadata, and digital signature

It enables:

• AI-native workflows
• Smaller files (~50–100 KB)
• Traceable & editable image generation

😓 Challenges I ran into

• Figuring out how to simulate a real latent vector
• Creating a readable file structure that could scale
• Encoding metadata and making it regenerable in tools like Stable Diffusion

🎉 Accomplishments I'm proud of

• Designing a fully new file format from scratch
• Making .AIGX 90–95% smaller than raster formats
• Embedding prompts and metadata for transparency and traceability

📚 What I learned

• How latent vectors, quantization, and compression work in AI pipelines
• How to build custom file containers using Python
• How to make AI more ethical, sustainable, and efficient

🔮 What's next for .AIGX

• Add encryption + creator verification
• improve to the general aspect of .AIGX
• Build a web viewer and editor
• Submit .AIGX as an open spec for generative AI platforms

Built With

• cryptography
• flask
• hashlib
• numpy
• pillow-(pil)
• python
• react
• rust
• streamlit
• transformers
• webassembly
• zipfile
• zstandard