Sign In

"One Image Is All You Need" Second Generation

24
136
508
5
Updated: Nov 26, 2023
characteranimewoman
Verified:
SafeTensor
Type
LoRA
Stats
136
508
Reviews
Uploaded
Nov 26, 2023
Base Model
SD 1.5
Training
Steps: 400
Epochs: 20
Usage Tips
Clip Skip: 1
Training Images
Download
Hash
AutoV2
F929728E79
0
0
0
0

This Lora is associated with the article One Image Is All You Need. Unlike the previously posted examples though, this Lora uses 6 images generated from a single image Lora (the red haired lady with a face mask) to reinforce the character and make a more flexible model. This was used to produce the second generation images in the article.

The source images were all produced using FoolCat 3D Cartoon Mix and this is the parent model you should use if you want to replicate the originals closely, but they should work reasonably well with any other model.

Resolution: 512,512 and 512,768
VAE: vae-ft-mse-840000-ema-pruned
Recommended strength: 0.4-0.6 (for epoch 20)

It's recommended to use these negative embeddings, but they are not necessary:

Captions: woman, long red hair, brown eyes, ear, earring, close up, white coat, black blouse, tie, bokeh, mountains, forest, lake, sky, clouds

Captions used only once: medical face mask, ears, tie clip, from side, hand on wooden rail