this post was submitted on 01 Oct 2023
269 points (100.0% liked)

196

16440 readers
1661 users here now

Be sure to follow the rule before you head out.

Rule: You must post before you leave.

^other^ ^rules^

founded 1 year ago
MODERATORS
 

Ai Generated image of a parking lot with a KFC that is mid-explosion and a chicken being launched by said explosion, image captured by what looks like a CCTV

you are viewing a single comment's thread
view the rest of the comments
[–] camr_on@lemmy.world 33 points 1 year ago (2 children)

The sign being correct is actually impressive. How many iterations did you go through?

[–] Mint@lemmy.one 22 points 1 year ago* (last edited 1 year ago) (2 children)

None for the text, Dall'e 3 for the most part tends to nail text, and I would say most of the time, it get text right that you specifically ask for.

So say if you ask for Pepsi most of the time the text will eligible with maybe some oddities at worst, but if you ask for something general like "Magazine cover" and not specify the text that should be on there, it will likely be unlegible.

Though in the case of Pepsi because it is a brand with ton of photos I assume its easier for it to nail, where as say if I asked for a sign with "x" text it out of the four images one or two will be legible, and the more text you will put the more likely for it to be nonsensical.

[–] solinus@lemmy.cafe 3 points 1 year ago

the advancement of technology never ceases to amaze and frighten me

[–] camr_on@lemmy.world 2 points 1 year ago

Really cool. I didn't expect ai to make so much progress on text in images so quickly. I'll have to go play around with it

Isn't dalle-3 actually good at text now?