-
Posts
1,099 -
Joined
-
Last visited
Silly Druid's Achievements
Single Status Update
See all updates by Silly Druid
-
Happy Easter!
Image source: purplesmart.ai
- Show previous comments 2 more
-
Yeah, that's how it works in theory, but these arts are so nice it makes me wonder if it's possible to tell the artists judging by the style in the ai generated piece or if it's actually so mixed, that it's something "new" per se. I mean, surely it mixes some techniques and styles it learned from so it might be tricky, but have people ever managed to guess what it based on making their piece or is it too unique in its own way?
Either way, I gave it a try
I can see what kind of prompts and negative prompts I should have added to guide it better, but I guess it generally delivered what I asked for.
can you guess the prompt?
SpoilerTwilight sparkle wearing a tophat, proudly standing in front of Golden Oak Library in the middle of the day, alone
I don't know where it learned, that Twilight has 3 hooves though. Also actual tophat-styled mane apparently
-
@Rikifive I haven't heard of things like guessing the original in case of an image with a text prompt.
Yeah, sometimes it messes up and even things like the number of limbs are not right. Generally you will have to make many attempts before you get something that looks really good. Also adding some things to the prompt (and negative prompt) that increase the overall quality of the image and direct its style where you want it to be is not a bad idea. For example, I add "plushie" to the negative prompt, because sometimes it can generate images that look like plushies, which kind of happened in your example. Unless you want it to look like a plushie, in that case add it to the positive prompt.
(Btw it seems the AI doesn't know how the Golden Oak Library looks like.)
-
And I definitely should have added something like "digital art / painting" in the prompt for starters, that alone *should* automatically guide it better to something I aim for, because it kinda went with realistic-photo-style, which is obviously far from what I expected to get.
And yeah, I imagined, that if it is a pony-themed tool, it will know better, but apparently one would have to specify, that it's the library from MLP in ponyville, maybe then it would get it right. Perhaps adding "My Little Pony" to the prompt would also help to make it focus more on MLP, not random real life stuff.
That was just my first attempt to see what the ai generates "by default", so to speak, and what I'll have to put to guide it. I assumed it will strongly stick to MLP by default, so I didn't put any such obvious tags, but apparently you do have mention you want it to related to MLP, and nothing real-world-looking..
- Show next comments 3 more