Nophica FFXIV

[vapid crest]

I can't seem to get any TI or LoRA to work from A1111 and various extensions. Could anyone make one for Nophica from FFXIV? I have already compiled ~60-70 x512 images.
https://drive.google.com/file/d/16qyCIsuLnyXLd2wilner6ejSMW_eFXC9/view?usp=sharing

else https://danbooru.donmai.us/posts?page=1&tags=nophica
https://gelbooru.com/index.php?page=post&s=list&tags=nophica
https://www.deviantart.com/search?q=nophica

If you could post settings used so I can try to replicate it would be appreciated too. Thanks!

[autumn violet]

I used only your set of images, I trained the embeddings during 6000 steps

here is the config file I used if you want to replicate the embedding

And the parameters for generating the example images :

1girl, best quality, masterpiece, highly detailed, nophica
Negative prompt: (worst quality, low quality:1.2), normal quality, greyscale, simple background, (jpeg artifacts:1.2), blurry, errors
Steps: 50, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 3711341399, Size: 512x512, Model hash: 038ba203d8 (AbyssOrangeMix2_sfw), Batch size: 4, Batch pos: 0, Denoising strength: 0.58, Clip skip: 2, ENSD: 31337, Hires upscale: 1.6, Hires upscaler: Latent

Unfortunately the training images are too compressed and the embedding generates bad looking images. I solved it by scaling up the generated image (using highres fix on A1111 webui), but still I am not satisfied with the embedding, this is why I didn't published it in the "ti-embeddings" channel. Please let me now if you are satisfied and if you want to publish it though.

[vapid crest]

Thank you very much! We don't need to publish it, I was really hoping to use this as a learning experience as nothing I try works. This and the yaml are a GREAT help, and the embedding doesn't look that bad either!

What trainer did you use? Once again, many thank yous! Exactly what I was hoping for.

[autumn violet]

Sorry I am french, I am not sure I understand the question "What trainer did you use" do you mean that ? https://colab.research.google.com/drive/1bbtGmH0XfQWzVKROhiIP8x5EAv6XuohJ

or do you mean the model I used to train the embedding ? nai-wd.ckpt