Feasibility of Adapting ART to Work with Other Models

[useruses001]

First, thank you for sharing this incredible research and implementation. The multi-layer transparent image generation capabilities are impressive.

Looking at the codebase, I noticed that ART uses FLUX.1-dev as the underlying model. I'm curious about the feasibility of adapting the ART framework to work with other foundation models like Stable Diffusion.

Specifically, I'd like to understand:

1. How tightly coupled is the ART implementation to FLUX.1-dev? Are there specific capabilities of FLUX that other models lack?

2. What would be the main challenges in adapting ART to work with Stable Diffusion or other open-source models?

3. Is the multi-layer transparent generation capability primarily implemented in your wrapper code or does it rely on specific modifications to FLUX?

4. Are there any plans to support other models in the future?

I appreciate any insights you can provide on this topic. Thank you for your time.