4 ms·

Fair enough. I dislike word-expansion AI text so I get it. When I write on my blog, part of it is as memory like a diary for myself and I find that I can create

• nemomarx 3mo ago
  The painting style does make these look less obvious than the default, so that helps. I think this is the best generated art I've seen so far. Although if it's for a mental diary maybe you could get a nice camera for it?
• novaleaf 3mo ago
  that's pretty nice illustration work, I'd agree, keep doing it :D
  • arjie 3mo ago
    I thought Nano Banana did a great job! I used a trick I saw someone describe elsewhere where I inserted the annotations on the photo / google street view of the scene. It seems like the model does a better job of instruction compliance when you annotate on the image itself rather than provide text. So an arrow to place the penny farthing guy and so on.
• Palomides 3mo ago
  it seems kind of perverse to to generate an image of a scene (fundamentally not derived from how it actually looked) to remember how it looked; you even immediately point out a way that it's inaccurate ("They looked younger and in better shape but you get the picture.")
  • arjie 3mo ago
    It's pretty neat, actually. If you use Gemini's Nano Banana with a photo reference, it's pretty good at keeping most of the things that matter. I walk down 4th Street daily and that's pretty much what it looks like near Berry St. And the thing I'm trying to remember is not, say, her athleisure or his white jacket. I'm trying to remember a couple with a dog with the man carrying a bag and a garbage grabber. This notion of precision is actually an interesting thing to me. It's a question of what I'm trying to memorialize. Their clothes or their fitness or the colour of their dog doesn't matter so much. You know, like some people object to statues of John Muir because he said some things like calling indigenous people "dirty and irregular" and stuff like that. But I think statues don't memorialize the man, they memorialize the symbolic man. The actual John Muir is dead, and we have his memetic descendant John Muir The Image. John Muir the image is a chalk drawing of John Muir the man, and the fact that the latter liked a dog, or hated boiled potatoes, or loved a woman, or whatever is irrelevant to me. I have taken this person, and flattened him on a canvas, and painted his acts on them as a guideline for myself. The rest of him is gone. And I like it that way. I've had this as a draft post on the blog for a while so if you don't quite get what I'm saying about this accepted imprecision, it's probably because I haven't yet found a way to write this idea out in a clearer way.
    • Palomides 3mo ago
      I think you would find learning to draw extremely rewarding
      • arjie 3mo ago
        Knowing how to, perhaps. Learning how to, sadly not. But my wife is an artist / creative director so there’s some adjacency there.
• abustamam 3mo ago
  For my wife's birthday, I generated a fantastical image of our family (me, her, daughter, two cats) because staging such a thing IRL would have been difficult. It took me several hours to take some nice photos of my cats and daughter and then composite them and then generate an image i liked. Then I sent it to a framer and got it framed. She loved it! I think there's a time and a place for AI generated images that actually require care to generate a good result. I guess to use a cliché, its the thought that counts.
  • arjie 3mo ago
    Haha, it's all in how you use the tool, eh? My wife actually used an early Dall E to iterate on the design for my wedding sherwani. I loved the end result too. Picture here: https://x.com/arjie/status/1855328068883353665?s=20 https://x.com/arjie/status/1855328068883353665?s=20