- First LoRA model and first time posting here, this model was trained on 120 hand picked images (cosplay, nsfw, sfw) of Shinobu Oshino from Monogatari Series. The base model for training was AbyssOrangeMix2, it works great with AOM3 and others, you should test it in different models and choose one you find better. Needs a decent prompt.
- Don't forget to favorite to get notified when there is any update.
- Recommended weight 0.5 to 0.8, and HiResFix. Also fix the LoRA file name if you use my prompt.
- Take a look on my prompts for the example images so you don't get lost on how to pull an style or a clothing.
- Note that I use XFORMERS and Negative TI's, so if you're trying to recreate my example images, it will be different.
- I could do a prune/resizing, but full size has better quality. (maybe if there's a need I can add it later)
- If you use it and post on Twitter or Pixiv, please tag me or in the comments, I want to see your creations! You all are free to comment, I appreciate any advice to make it better. Thanks!
Description
First version of this model, beware of errors. Trained on AbyssOrangeMix2, with 120 hand picked images of Oshino Shinobu, 3200 steps and 4 epochs.
FAQ
Comments (14)
- New images uploaded, the older images we're all removed because of a tag.
hello, very good, your loRA wanted to know how you do it, do you have a link where you learned from, thank you very much
Follow any guide on Youtube of how to train an LoRA model, just to get familiar and how to install it, then I recommend to follow this guide https://rentry.org/lora_train. Note that I didn't followed this exact steps, I did some more research about it, but this is a good starting point 👍
@zaadsatan thank you very much you are my idol
@Mudo No problem at all XD, follow me and leave a positive review if you want also ✌😊
what checkpoint models you use to generate the first, fourth, fifth, sixth and eighth pictures?
All of them look incredibly good, I really want to know how you made them
AbyssOrangeMix3
@zaadsatan I try my best but can't get anything close to them, sad. What's Negative Ti's?
@electrochemistr It's negative embeddings or negative TI's for short (Textual Inversions), it helps to get better results! You can filter to only show Textual Inversions here on civitai and search for good negative TI's, so you can put on the negative prompt 😊
This one is a good starting point if you don't have any https://huggingface.co/datasets/gsdf/EasyNegative
Thank you
I love u, m8 <3
Thanks so much for sharing your excellent talent with the community. It’s really great to see my waifu in action like this.
Absolute Legend! Thanks
crab girl
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.