6 ms·

Maybe I'm an obscure case, but I'm just not sure what I'd use an image generation model for. For people that use them (regularly or not), what do you use them

Maybe I'm an obscure case, but I'm just not sure what I'd use an image generation model for. For people that use them (regularly or not), what do you use them for?

vunderba 11mo ago
Mostly highly specific images in blog posts but I also use it for occasional comics. https://mordenstar.com/portfolio/gorgonzo https://mordenstar.com/portfolio/gorgonzo https://mordenstar.com/portfolio/brawny-tortillas https://mordenstar.com/portfolio/brawny-tortillas https://mordenstar.com/portfolio/ms-frizzle-lava https://mordenstar.com/portfolio/ms-frizzle-lava
- jasonjmcghee 11mo ago
  I'm kind of reading between the lines, but sounds like "for fun" which makes sense / what I generally expected for why people use it
  - vunderba 11mo ago
    I think that's a fair assessment. I write a lot of bizarre fiction in my spare time, so Text2Image tools are a fun way to see my visions visualized. Like this one: A piano where the keyboard is wrapped in a circular interface surrounding a drummer's stool connected to a motor that spins the seat, with a foot-operated pedal to control rotation speed for endless glissandos.
cj 11mo ago
Random examples: 1) I have a tricep tendon injury and ChatGPT wants me to check my tricep reflex. I have no idea where on the elbow you're supposed to tap to trigger the reflex. 2) I'm measuring my body fat using skin fold calipers. Show me were the measurement sites are. 3) I'm going hiking. Remind me how to identify poison ivy and dangerous snakes. 4) What would I look like with a buzz cut?
- jasonjmcghee 11mo ago
  First three are interesting - all question / knowledge based where the answer is a picture. Hadn't really considered this.
  - mrguyorama 11mo ago
    The answer is a picture that almost certainly already exists. Why would you want a program that just makes one up instead?
    - phatfish 11mo ago
      So you can feel 1000x better about yourself when 1000x more resources are used to create an extra special image just for you. Rather than the canonical one served from the Wikipedia (or Google image search) cache.
- paulglx 11mo ago
  You should never rely on AI to do 1, 2 or 3, especially a sloppy model like this.
hemloc_io 11mo ago
porn is probably the a biggest one? but concept art, try-it-on for clothes or paint, stock art, etc
xnx 11mo ago
Nano Banana is more of an image editing model, which probably has more broad use cases for non-generative applications: interior decorating, architecture, picking wardrobes, etc.
- jasonjmcghee 11mo ago
  Yeah... For some reason none of these are use cases in my day to day life. That said, I also don't open Photoshop very often. And maybe that's what this is meant to replace.
  - xnx 11mo ago
    Not for everyone everyday, but a good tool to have in the toolbox. I recently was very easily able to mock up what a certain Christmas decoration would look like on the house. By next year, I'm sure that feature will be part of the product page.
- vunderba 11mo ago
  Definitely, but don't sleep on its generative capacities either. You can give it a image and instruct it "Use the attached image purely as a stylistic reference" and then proceed to use it as a regular generative model.
  - xnx 11mo ago
    Indeed. Is Nano Banana now Google flagship image gen model (over Imagen 4)?
    - vunderba 11mo ago
      In my tests it does outscore Imagen3 and Imagen4 even in the generative capacity, but my benchmark is more focused around prompt adherence. I'd wager that for certain photorealistic tests Imagen4 is probably better. https://genai-showdown.specr.net/?models=i3,i4,nb https://genai-showdown.specr.net/?models=i3,i4,nb
esafak 11mo ago
I'm creating a team T-shirt from a bunch of kids drawings. The model has synthesize a bunch of disparate drawings into a cohesive concept, incorporate the team's name in the appropriate color and font, and make it simple enough for a T-shirt.
TheAceOfHearts 11mo ago
My most regular use-case is generating silly memes in group chats. If someone posts something meme-worthy or I come up with a creative response, image generation is good for one-off throwaway memes. A recent example was an "official license to opine on sociology", following someone arguing about credentialism. Recently I also started using image generation models to explore ideas for what changes to make in my paintings. Although generally I don't like the suggestions it makes, sometimes it provides me with creative ideas of techniques that are worth experimenting with. One way to approach thinking about it is that it's good for exploring permutations in an idea-space.
hooverd 11mo ago
Nonconsenual pornography is the killer app.

Maybe I'm an obscure case, but I'm just not sure what I'd use an image generation model for. For people that use them (regularly or not), what do you use them for?

vunderba 11mo ago

Mostly highly specific images in blog posts but I also use it for occasional comics. https://mordenstar.com/portfolio/gorgonzo https://mordenstar.com/portfolio/gorgonzo https://mordenstar.com/portfolio/brawny-tortillas https://mordenstar.com/portfolio/brawny-tortillas https://mordenstar.com/portfolio/ms-frizzle-lava https://mordenstar.com/portfolio/ms-frizzle-lava

jasonjmcghee 11mo ago
I'm kind of reading between the lines, but sounds like "for fun" which makes sense / what I generally expected for why people use it
- vunderba 11mo ago
  I think that's a fair assessment. I write a lot of bizarre fiction in my spare time, so Text2Image tools are a fun way to see my visions visualized. Like this one: A piano where the keyboard is wrapped in a circular interface surrounding a drummer's stool connected to a motor that spins the seat, with a foot-operated pedal to control rotation speed for endless glissandos.

cj 11mo ago

Random examples: 1) I have a tricep tendon injury and ChatGPT wants me to check my tricep reflex. I have no idea where on the elbow you're supposed to tap to trigger the reflex. 2) I'm measuring my body fat using skin fold calipers. Show me were the measurement sites are. 3) I'm going hiking. Remind me how to identify poison ivy and dangerous snakes. 4) What would I look like with a buzz cut?

jasonjmcghee 11mo ago
First three are interesting - all question / knowledge based where the answer is a picture. Hadn't really considered this.
- mrguyorama 11mo ago
  The answer is a picture that almost certainly already exists. Why would you want a program that just makes one up instead?
  - phatfish 11mo ago
    So you can feel 1000x better about yourself when 1000x more resources are used to create an extra special image just for you. Rather than the canonical one served from the Wikipedia (or Google image search) cache.
paulglx 11mo ago
You should never rely on AI to do 1, 2 or 3, especially a sloppy model like this.

hemloc_io 11mo ago

porn is probably the a biggest one? but concept art, try-it-on for clothes or paint, stock art, etc

xnx 11mo ago

Nano Banana is more of an image editing model, which probably has more broad use cases for non-generative applications: interior decorating, architecture, picking wardrobes, etc.

jasonjmcghee 11mo ago
Yeah... For some reason none of these are use cases in my day to day life. That said, I also don't open Photoshop very often. And maybe that's what this is meant to replace.
- xnx 11mo ago
  Not for everyone everyday, but a good tool to have in the toolbox. I recently was very easily able to mock up what a certain Christmas decoration would look like on the house. By next year, I'm sure that feature will be part of the product page.
vunderba 11mo ago
Definitely, but don't sleep on its generative capacities either. You can give it a image and instruct it "Use the attached image purely as a stylistic reference" and then proceed to use it as a regular generative model.
- xnx 11mo ago
  Indeed. Is Nano Banana now Google flagship image gen model (over Imagen 4)?
  - vunderba 11mo ago
    In my tests it does outscore Imagen3 and Imagen4 even in the generative capacity, but my benchmark is more focused around prompt adherence. I'd wager that for certain photorealistic tests Imagen4 is probably better. https://genai-showdown.specr.net/?models=i3,i4,nb https://genai-showdown.specr.net/?models=i3,i4,nb

esafak 11mo ago

I'm creating a team T-shirt from a bunch of kids drawings. The model has synthesize a bunch of disparate drawings into a cohesive concept, incorporate the team's name in the appropriate color and font, and make it simple enough for a T-shirt.

TheAceOfHearts 11mo ago

My most regular use-case is generating silly memes in group chats. If someone posts something meme-worthy or I come up with a creative response, image generation is good for one-off throwaway memes. A recent example was an "official license to opine on sociology", following someone arguing about credentialism. Recently I also started using image generation models to explore ideas for what changes to make in my paintings. Although generally I don't like the suggestions it makes, sometimes it provides me with creative ideas of techniques that are worth experimenting with. One way to approach thinking about it is that it's good for exploring permutations in an idea-space.

hooverd 11mo ago

Nonconsenual pornography is the killer app.