Sign In

Inkpunk768

1.6k
11k
27
Verified:
PickleTensor
Type
Embedding
Stats
8,371
Reviews
Uploaded
Dec 12, 2022
Base Model
SD 2.0 768
Training
Steps: 2,500
Trigger Words
InkPunk768
Training Images
Download
Hash
AutoV2
F96DDF8E51
0
0
0
0

These embeddings are trained on output from the Inkpunkv2 1.5 custom checkpoint. I really like the InkPunk style, but I wanted to be able to mix and match it with other high quality 2.0/2.1 embeddings, such as myΒ SDA768 andΒ GTA768Β embeddings.

--------

To use these embeddings in AUTOMATIC1111, copy them to your embeddings folder in your stable-diffusion-webui directory. No need to restart, just invoke the embedding from your prompt. Make sure you are using a compatible model or you will get errors!

--------

These InkPunk embeddings were all trained on 51 768x768 images that were cleaned up and labeled for training. Training was performed at 0.0005 learning rate, stopped at 3500 steps as it was starting to overtrain by that point.

I've included 4 different embeddings to use:

InkPunkHeavy768 - This will lay down thick lines and cross hatching, but still give a beautiful inked and sketched style.

InkPunk768 - This one outputs closest to the InkPunk custom checkpoint in my opinion, you'll get a nice mixture of hand drawn and painted look

InkPunkLite768 - If you're finding that you're losing too much detail to the hand drawn look, this light embedding may bring back some of the detail you're looking for.

InkPunkLandscapes768 - This 1200 step trained embedding doesn't look great for people, but it sure can make some pretty buildings and landscapes in the hand inked style.

All of these embeddings require model version 2.0 and up. They are all trained on 768x768 images, and use a total of 8 vectors. They will work just fine with the 2.0 512 models as well, including the depth mapping model (in img2img mode)


πŸ’–πŸ’–I hope you enjoy using these embeddings in your art as much as I enjoy creating them for the Stable Diffusion community. πŸ™πŸΌπŸ™πŸΌIf you do like them, can you pretty please click the heart at the top and leave a review? Many thanks! πŸ’–πŸ’–