Description
FAQ
Comments (4)
Changed the settings when making the model. Not sure its that much better. Its so hard to tell. The first one I didn't use blip and 700 steps with Gradient accumulation steps set to 3 and batch size of 4, learning rate .004. This one everything is set to default and I used blip trained to 3400 steps. Used the inspect script to make sure it wasn't overly trained.
Well done job for version 2.
It deserves a star rating.
.
Here is a suggestion for you, learn from a model-creator named JernauGurgeh.
He also does alot of T.I. celeb models with VERY Great results.
Search for " Eva Longoria" (his recent model) so you can find his posts.
He also posts HOW he does his training and he has discovered the best way to train a model for a
Textual Inversion. He also gives out many tips and tricks in his post-answers when answering
comments on his models. So, also look into them.
If you are able to create an extremely good model the way he does, then you will soon become
a master at using Stable-Diffusion.
Cheers
MAX
@maxhitman Thanks Max. I will look into it.
@maxhitman I been trying to make models with less images and find the body is all wrong. The model makes the people too tall sometimes, just feel it don't really look like their body. It works great for close ups, or if you just want a hot woman and put everything in the prompt. He got a good method for what he trying to do, Glam shots.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.