In Bing Image Creator, Dal•e 3
Prompt: a knit ugly Christmas sweater celebrating mechanical engineering, featuring gears, pistons, wheels, diagrams, and the phrases “DRAW STUFF” and “BUILD STUFF!”
Prompt: a chaotic ugly Christmas sweater, covered in things related to mechanical engineering. it is a knit sweater. it includes the words “draw stuff” and “build stuff”
Prompt: an ugly Christmas sweater, covered in knolled assortment of things related to mechanical engineering. it is a knit sweater. it includes the words “draw stuff” and “BUILD STUFF!”
How do you get the AI to reliably do text like that? Every time I try and get it to do text it comes out garbled.
I’m using Bing Image Creator, with Dal•e 3. It’s not really reliable, the results above are the best pictures out of 10 prompts with tweaks each time, and there are still some mixed up or repeated letters up there.
Here are some examples with worse messed up text:
Oh that makes sense. I haven’t played with Dalle 3 yet. I know that with stable defusion and even SDXL and it’s almost impossible to get it to do coherent words. I’ll have to mess around with it later
Buld stuff is gonna be my new Github bio.
I think OP is using dalle 3. (Bing image generator). It has really good text handling capabilities and will abude by the prompt to some extreme degrees.
I just tried with the exact same prompt and got a pretty nice sweater except the text says “UVLY CHRTISTMAAS” and “ULBLY CHRIISTMAS”
Oh no! Is it still working with Dal•e 2?
I’m not sure! I suck at getting decent results. I don’t really know what I’m doing with image generation… especially with Bing
Confirming Dal•e 3