3 ms·
To be fair, it is not always that awful. Here is a sample of the results of simpler subset prompts I like to run on image generation: https://imgur.com/a/aO5S7y
by EvgeniyZh 3y ago
To be fair, it is not always that awful.
Here is a sample of the results of simpler subset prompts I like to run on image generation:
https://imgur.com/a/aO5S7yM https://imgur.com/a/aO5S7yM
Some are bad (first two), but others are okay; it understands text pretty well, but the artifacts just feel like years ago.
I still can't understand how it got released and advertized.
- kossTKR 3y agoThanks for the examples. I've used SD / Midjourney / Dalle extensively and would say this is honestly shockingly bad besides the two last ones. Comparable to the first versions of the other services, but with better contextual understanding but still with lots of gnarly artifacts and weirdness going on.