Eh, it used to be the case that some models generated images with watermarks, as those pics were part of their training data. I remember seeing some hard to read alarmy and Shutterstock watermarks on ai posts
Sometimes yes. Those old models saw that exact text a LOT so yeah. Just like how today, stable diffusion XL can somewhat reliably generate glasses of Nutella.
Eh, it used to be the case that some models generated images with watermarks, as those pics were part of their training data. I remember seeing some hard to read alarmy and Shutterstock watermarks on ai posts
they were legible??
Sometimes yes. Those old models saw that exact text a LOT so yeah. Just like how today, stable diffusion XL can somewhat reliably generate glasses of Nutella.
damn