Merzmensch Kosmopol🧑‍🎨🤖さんのプロフィール画像

Merzmensch Kosmopol🧑‍🎨🤖さんのイラストまとめ


Artist, Writer, Data Journalist #Art #Culture #Dada #Avantgarde #Merzkunst #Japan #Transmedia #AI #VR #IndieGames Generative AI 🇯🇵🇩🇪 | With AI since 2016
merzmensch.com

フォロー数:10780 フォロワー数:11438

Prompt "A man in a liminal space between past and future, looking for the meaning of life and speaking with the entire world"

0 2

Prompt "Japanese Futurism" doesn't look Japanese, but futuristic at least.

I wonder why we see here a typical village wooden fence with some tables.

District 9 vibe, definitely.

V4

0 4

Indeed! img2img is THE THING.
Here is my original (1) and three img2img results. THIS is gamechanger. https://t.co/gdzHxnzJxR

7 87

Yes, actually that's the thing! Gen models create specific stuff if they don't get it. DALL·E does nature/still life photos.

Midjourney V3 did impressive mix of surrealism / expressionism (left). V4 (right) creates girl portraits if it doesn't get the prompt

📝fhrgygbtfty

0 6

I dig this "Dystopian Elephant" (my prompt) by MidJourney V4

P.S. But DALL-E is cool as well (right)

0 8

Above images are without "Chaos" as feature.
And now if we add "--Chaos" with factor "100", we get really different results:

1 5

OK, I admit it's cherry picked, the other 3 are elephants, but nevertheless:

0 5

re: content overload in MidJourney V4
(don't get it wrong, I LOVE the quality of MidJourney, just some constructive observations):

Prompt: 🐘

Yes, just an emoji of an elephant, not more or less.

Completion =>

I mean, I see an elephant, but a woman and butterfly?

0 4

Left: "kitsch"
Right: "anti-kitsch"

Imho both: kitsch

1 1

Version 4 does too much. Even if visually striking, it looks like photoshop surreal art from earlier 2000ies. Too many details are not always good. It's like an artist who isn't satisfied with his work and tries to enhance it into kitschy ways. Again,it's about prompts too.

1 18