I have generated some comparative images between the old and the new model.
some clarifications:
I used the same “generic” prompt for both models. Surely the quality can improve a bit by using a prompt optimized for the specific model but I wanted to keep the test as neutral as possible.
no cherry picks: these are the images that came out, I did not discard and kept only the best ones.
in general both models follow the prompt fairly well with sometimes a slight advantage of one over the other and vice versa.
the details. don't just look at the girl but also look at everything else, the furniture in the background, the snow, the sea, the waves, the sky, the clouds, the clothes, the fabrics etc etc.
On this I think I cannot be contradicted by saying that the new model wins without a doubt.
as far as I am concerned I have no doubt, the new one is definitely superior to the old one but if you feel it necessary I can use both alternating them, we just have to figure out in which contexts to use one rather than the other.
PeteDeeper
2024-12-08 23:07:28 +0000 UTCghostbeetle
2024-12-07 08:22:04 +0000 UTCRick
2024-12-06 21:59:59 +0000 UTCBJax
2024-12-06 19:11:40 +0000 UTCSpinal121
2024-12-06 18:31:55 +0000 UTCJeffarooney
2024-12-06 18:05:44 +0000 UTC