Murkrow
Says "also" and "or something" a lot
- Pronoun
- he
DallE mini has been doing the rounds on twitter recently and I've been spamming discord servers with funny images it's generated for me.
Another ai image generation site I've had fun with in the past is NeuralBlender, though the free version of that seems to have got worse since I last played it with late last year.
Share some of creations/nightmares you've generated here! All in one place so you don't have to spam it anywhere else :P Feel free to share things other than images as well, like ai-generated text.
Here's some of mine:
DallE is pretty good at making images of Walter White - I expect it's because he's a famous character and many images of him are closeups of his face. I think the top-middle Sugimori one is my favourite
"Trainer Cynthia" didn't work as it just came up with pictures of women with horses. I think because the word Pokémon is in the prompt is has tried to make some of the images look more like a Pokémon species, like the centre one looks a bit Gardevoirish. The bottom-right one looks like a cross between Cynthia, Mawile, and Muttley.
I very much like the range of facial expressions you get as well.
Top left looks deep in thought or annoyed
Middle-left looks a n g e r y
Bottom-left looks like she doesn't care - talk to the hand!
Bottom-middle looks like she's tutting or pouting.
The AI really good at generating pictures of Funko Pops, provided it knows the character you enter. It's probably because the way Funko Pops are a bit uncanny valley in the same way that the AI generates images (dead eyes, simpler shapes are easier, etc.)
I thought Gardevoir is somewhat humanshaped and is an unambiguously named character - it came out more terrifying than I expected.
Another character who is quite unambiguous who you mean. Except this time it's a character where I expect there's orders of magnitude more fanart than official art, and that definitely shows in that these images are less anime and more of a tumblr/deviantart kind of style.That, and the fact that it seems to want to make her chest big
One thing I've notices when generating images of characters wearing uniforms is that there isn't much variance between uniforms within the same image - in this example they're almost always all the same colour. Even when it generates characters who are supposed to be in the uniform it often gets the colours wrong. I think that's super interesting and offers a glimpse into the AI's "thought" process.
Another ai image generation site I've had fun with in the past is NeuralBlender, though the free version of that seems to have got worse since I last played it with late last year.
Share some of creations/nightmares you've generated here! All in one place so you don't have to spam it anywhere else :P Feel free to share things other than images as well, like ai-generated text.
Here's some of mine:
DallE is pretty good at making images of Walter White - I expect it's because he's a famous character and many images of him are closeups of his face. I think the top-middle Sugimori one is my favourite
"Trainer Cynthia" didn't work as it just came up with pictures of women with horses. I think because the word Pokémon is in the prompt is has tried to make some of the images look more like a Pokémon species, like the centre one looks a bit Gardevoirish. The bottom-right one looks like a cross between Cynthia, Mawile, and Muttley.
I very much like the range of facial expressions you get as well.
Top left looks deep in thought or annoyed
Middle-left looks a n g e r y
Bottom-left looks like she doesn't care - talk to the hand!
Bottom-middle looks like she's tutting or pouting.
The AI really good at generating pictures of Funko Pops, provided it knows the character you enter. It's probably because the way Funko Pops are a bit uncanny valley in the same way that the AI generates images (dead eyes, simpler shapes are easier, etc.)
I thought Gardevoir is somewhat humanshaped and is an unambiguously named character - it came out more terrifying than I expected.
Another character who is quite unambiguous who you mean. Except this time it's a character where I expect there's orders of magnitude more fanart than official art, and that definitely shows in that these images are less anime and more of a tumblr/deviantart kind of style.
One thing I've notices when generating images of characters wearing uniforms is that there isn't much variance between uniforms within the same image - in this example they're almost always all the same colour. Even when it generates characters who are supposed to be in the uniform it often gets the colours wrong. I think that's super interesting and offers a glimpse into the AI's "thought" process.
