Raw LLM Responses

20:42 I think I've seen enough. Going into this video with your introduction, I thought "Wow, a 1.5 hour long video with a guy who knows how stable diffusion works". But by now, I'm a little disappointed. Look, I cannot agree with Shadiversity's claims of the resulting work being entirely his own. I think it's pretentious and highly disrespectful to artists. The resulting work is an AI-generated composite at best. However, I now question how well you understand the diffusion process to make the claim that this technique could be fixed with "better prompting". You might not like hearing it, but for what he does, he does it well. His technique is solid, inpainting and denoising is a skill. An artistic skill? Debatable, but a skill none the less. Stable Diffusion is clever, but it can't read your mind. This is a limitation of how text encoding works: When you make any modification to the prompt, its effect compounds on a per step basis, often changing other aspects of the image unrelated to what you requested. If he had simply "done the right prompting", the image's result could change drastically. And to be completely fair to Shadiversity, this was far closer to him improving his art skills than just purely relying on prompting. You seem like a smart dude - and generally speaking, I agree with your notion that AI generative content is not true art. But I suggest revisiting this take after some more research and experimentation, because I suspect you've got better philosophical arguments up your sleeve regarding this subject matter to win people over in the middle.