I've updated my popular TheAlly's Mix! It still retains its' solid good-solid-all-round-general-purpose qualities, but now has an added bit of realism (using the token photo).
I've used the same sample images as my first mix, so you can go and compare how this stacks up against the OG.
It's a mix of Waifu Diffusion 1.3 and Stable Diffusion 1.5, with a further Sigmoid Interpolated NovelAI.
Please consider joining my Patreon! Advanced SD tutorials, prompt reveals, explanations, exclusive embeddings, adult-art, from a female content creator (me!) patreon.com/theally
There are no trigger words required. The model "listens" to Danbooru tags, obeys prompts, and accepts Textual Inversions well!
Description
FAQ
Comments (16)
I love that you used the same prompts as your OG Ally Mix for easy comparison. Looking forward to trying this out 馃ぉ
I'm excited to see your images!
I'm downloading now. Can't wait to get my hands on this. Your images are killer!
Please tell me how to install SafeTensor files? I haven't found any normal information anywhere on the Internet. Or add this model in the usual format. The model looks great from the photo preview
if you run automatic1111, you install it the same way you do any other checkpoint. Its just a format that prevents you from potentially being pickled (malicious code) from a normal checkpoint file.
Like gunn187 said, just drop the file alongside your other checkpoint models and you'll be good to go! Thanks :)
Just a quick note for fellow DiffusionBee users: You'll need to convert this safetensor to a ckpt file and keep your fingers crossed that it doesn't error out when importing.
the thumbnail girl looks like Haley Bennett :O
Huh, I had to google her and you're right! I can see the resemblance :)
Tried recreating the fisherman prompt, but I can't seem to find the right hypernetwork. Any ideas where I can grab that from? I seem to be getting continuously bad anatomy or two headed fisherman... but not in the right way.
why many peopel use SD1.5 instead of 2.0?
A number of reasons! Because SD 2.0 doesn't have any naked ladies, mostly :) Also because it's harder to work with, harder to prompt, harder to train against. I'm sure that will change in future, but right now it's just not as versatile as 1.5.
@theally聽also SD2 doesn't recognize artist names and celebrity names. Harder to control faces and styles.
Careful, Waifu Diffusion 1.3 had problems with the dataset, so the booru tags it accepts require underscores. I would advise to find an alternative to that.
What VAE best fits the model?
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.