Trained on 19 Screencaps at 609 steps from one episode, then upscaled with Topaz Photo AI.
Hints on usage: "PonyXLV6_Scores" is an embedding that is linked in recommended resources.
USE AT THE BEGINNING OF PROMPT:
PonyXLV6_Scores, (anime coloring, 2d, anime_screencap), source_anime, 2d, original anime style,
frill_shiranui, long hair, green eyes, black hair, mole, green hair, mole under mouth, mole under eye,
LoRa Strength: (I only used 0.8)
Description
FAQ
Comments (2)
Thanks, nice LoRA ❤️
What is the checkpoint you are using and what are these embeds in your negative?
FastNegativeEmbedding-pdxl, negative_hand-neg-pdxl, bad-artist-anime-pdxl, bad-artist-pdxl, bad-hands-neg-pdxl, boring_sdxl_v1
I'm asking because I can't replicate the nice colors you have in your first image. The upscaling (and denoising as it seems since the screencaps are quite noisy) really push the quality imo :)
Also one correction: According to the model metadata you used 19 images and the step count is 608.
oh wow haha I missed that haha, yeah it was 19 images haha, I must have copies and pasted it from my other model page haha.
the checkpoint is my own model that I edited to have a stronger anime style of it's own.
and yeahh this was old, you can remove the embeddings, I found a embedding converter online for them to work in SDXL, pt to safetensors it would do.