/vt/ - Virtual Youtubers

>>71734164

>>71734443
Its a tough call anon.
Ignore the La+ quality here, that can be inpainted.
I feel like the animagine 3.1 understands and follows the tags very well. But kinda follows it to the T.
From this I can tell that it CAN make great background images. but it feels abit RNG for it to prompt well. for example, if my setting were to put bed, bedroom , lamp. The background will only have those 3 features most of the time. the randomness to it is lesser than that of 3.0.