Raw LLM Responses

Is it really the same as tracing? What if you asked an artist to do a piece similar in style to "insert reference image here"? I have seen it appearing in example images when looking for interesting data sets' example images, and I am a bit unsure about the effectiveness... For it to really work, there would have to be art of that specific individual, and they would have to be tagged with the prompt keyword to even be considered in this way. To make it even more complex, my understanding of how it works, is that if you say artist name: "StainedCoffeecup" - but there is nothing tagged with it, the prompt will gradually remove letters until the word becomes something sensible. In this case, my understanding is that you could risk getting results that then contain "more" stainedcoffeecup, which could show up as stains, coffee and cups. Most likely cups of coffee, and coffee stains. Unless you are a super famous artist, I also doubt the effectiveness of including artist style tags. Again, it would require that artists' work to be well represented in the data set. In other words, lots and lots of imagery. A smaller set like a lora might be more fitting of course, but that would also be easier to crack down on, as it would be pretty clear if the model author used those images to train data without asking first. And even then, if the artist produces wildly different style artwork in their published work instead of sticking to one, it would also be really difficult to replicate by an AI. So yea, I don't really get why people insist on a specific artist, unless they have a dedicated lora for it. Heck, some artwork is probably tagged such as :"authoername_inspired", so if you include a specific artist, it may look like the artist you want... but with a little bit extra of other artists that were inspired by it. But, I guess it can be fun to play with, if you were curious about what-if scenarios, such as a particular artwork re-imagined in the artstyle of specific artist. As an example with music and generative AI, I think it was quite fun to hear "gangsta's paradise", but sung by an AI Frank Sinatra. Sure, you could also have a singer do an impression of Frank, but just like generative AI, that wont ever be 100% perfect.