Analysis shows that indiscriminately training generative artificial intelligence on real and generated content, usually done by scraping data from the Internet, can lead to a collapse in the ability of the models to generate diverse high-quality output.
img2img is not “training” the model. Completely different process.
Oh yeah, you’re right. It’s both degradation in some way, but through entirely different causes.