I am pro open source. But it doesn’t make sense to open source slop or things that become very risky if not properly setup (training pipelines). It is incredibly difficult to automate dataset collection securely, so more human in the loop makes sense. With ai art, you can just use synthetic data. Create a base of what you’re going toward and train a LoRA. Then use that to generate more precise, custom, iterations. You never pull from online data sources. It’s probably best to do the same elsewhere