385
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 03 Mar 2024
385 points (97.3% liked)
Politics
5854 readers
6 users here now
Discuss world politics here.
Rules
-
Be civil and respectful
-
No NSFW content
-
No blatantly fake news or unfounded conspiracy theories
-
No heavily sensationalized articles
Community icon by Webalys, licensed under CC BY 3.0.
founded 4 years ago
MODERATORS
If you told me that was an AI generated image i would have sighed with relief.
I'm not convinced it's not - there are loads of potential artifacts, like around the blonde woman's eyes.
There's no way it'd get all of the text on the shirts and signs right
I don't expect everyone to know the different features of a "stable diffusion" style AI, but you can easily do something called img2img or inpainting, where you use transformations on a base image to help generate something more realistic. The current state of the art can easily generate readable text.
Hands. Those hands are too real.
Current state of the art generative AI doesn't have trouble with hands.
I use both mid journey and Dalle for work and that's untrue. Even the recent SORA has trouble with hands and it's the latest AI software.
Please tell me which AI generator software that doesn't have trouble with hands and I'll ask my company to switch to it.
If you're using ready-made solutions like MJ/Dalle, then we already know you're not at the state of the art, only at the state of business solutions...
Even something as straightforward as a local SDXL model with contextual inputs and a refining pass will get you pretty good results.
That doesn't sound very much "the current state". More like cutting edge.