Raw LLM Responses

not defending this character , however your ego tripping . First your explanation of how a neural network has trained and now stores the steps to denoising is lacking , because of the limited size of a model it cant store any of the original works instead it stores something like abstract functions that are conditionally activated . for example if your doing a digital painting of a landscape mountain and you want to add a lake you learn the "technique" of mirroring the mountain to create the reflection on the water , that is a simple function a geometric transformation that approximates the physical effect , we learn thousands of these "techniques" and this is probably the closest to what a neural network is actually storing ( conditional transformations ) , if you think that is stealing then it is a large stretch , infringement is something different if i draw someone elses art in their style thats infringement , but even then nothing was stolen . Secondly you denigrate the role of an artists skill to direct , Art Directors are an incredibly important role guiding the style and maintaining the tone of a project , but like prompting you seem to claim that being able to firstly recognize then articulate and guide the art towards the favored style is "not useful art skill" . Third your understanding of Context based editing is incorrect , he is using stable diffusion img2img in what looks like auto1111 webui , Context lora edits are a relatively new feature , something like Flux Kontext , or gpt's image generator , and they are lacking in the ability to control and get precise as edits as their making , his workflow is completely fine to make basic edits then use for img2img guidance but your talking trash when he is actually making photoshop painting edits about him not using Kontext in the way you think it works?