81–86 of 86 posts
Earlier quoted context omitted.
Since it’s legal (via explicit non-commercial data mining exemptions in EU law), it literally isn’t stolen.
It's not legal to reproduce copyrighted content. Idgaf where you live.
Earlier quoted context omitted.
I think the new images look jarring next to the original 8-bit Konami style font at the bottom, the look clashes in my mind. I would have pixelated the generated images and dithered them down to a smaller palette to look more retro. Or kept the hi-res images but subtituted the fonts to something more detailed and modern. But that's just me.
It reminds me of those HD texture packs for Minecraft in the early days that would attempt to map photorealistic textures onto a world almost entirely made of cubes. Yes it's more high-fidelity, but it adds nothing except emphasizing the low fidelity of everything else.
Earlier quoted context omitted.
Since it’s legal (via explicit non-commercial data mining exemptions in EU law), it literally isn’t stolen.
It's not legal to reproduce copyrighted content. Idgaf where you live.
And civil copyright violations aren’t “illegal” anyway.
Earlier quoted context omitted.
I still struggle with SD 2.0 prompting because many of the tricks (greg, artstation) don't work anymore, have people had success with it or do I have to use custom models?
You need other tricks and put a bit more thoughts in the negative prompt. The quality is then higher but you really need a good prompt. It’s a bit frustrating compared to midjourney or dall.e. You should take a look at embeddings too. They are tiny files, no more than 128kB, that have a huge influence on the final output. You put the files in the embeddings folder and use the filename in your prompt. Ideally the file…
I've seen this stated but it has not been my experience. Nor have I seen solid examples of it. For 2.0 (haven't tried 2.1) I find the model very finicky and unstable. Any prompt which works well also seems to work at least as well in 1.5.
Is there a way to have current AI tools maintain consistency when generating multiple images of a specific creature or object? For example, if there are images of 'Dr. Venom' they need to look similar, or if there are images of the same space ship.
I'd recommend giving it a shot if you have an Nvidia GPU with ≥4GB VRAM.
Edit: There are also training and hypernetworks, but they require a body of source material, keywording, and significantly more time and compute resources, so I haven't attempted either.
This is a great example of AI image generation being unable to generate "art" and instead just replicating a naive approximation of what it was trained on. There's no coherence or consistency between the images and while they all look "shiny", they also look incredibly dull and generic. It's a cool exercise but using this for a real-world project would eliminate any attempt at producing an artistic "voice". AI image…
Also, one model doesn't speak for them all. I have the problem that my results are often too consistent with certain models, largely because of prompt complexity, lack of wildcarding, etc.