-This is my Second XL Chara Lora -
Base model : AnyIllustrious-XL(for LoRA training)
Recommend a weight : 0.7 ~ 0.9
Recommend Models : WAI-NSFW-illustrious-SDXL,WAI-SHUFFLE-NOOB,etc...
Recommend Prompts (Chara) : kcl,1girl,
Recommend Prompts (Option): ponytail, pubic tattoo,
Recommend Prompts (clothing): cropped blue jacket, thighhighs, thigh boots, gloves, garter straps, thong, underwear, footwear,
Recommend Prompts (Negative): simple background, white background,
(Try different things if necessary.)
※Since the tag "kcl" aggregates tag elements such as hair style, hair color, etc, you probably do not need to bother setting it.
If it is less effective, try adding an optional prompt.
-Review(for personal reflection)-
After making various adjustments to the study material, I feel that some of it can now be output in Illustrious Checkpoint.
There is a limit to what I can try by myself alone, so I would be happy to have everyone try various things.
Please try different compositions. Adjust the intensity of Lora and enjoy!
Note : This is an experimental Lora to check its versatility, I may create a new version to increase its versatility.(Mainly about learning materials and tags)
I would like to see the work of everyone using this Lora, whether normal or NSFW, so please post a variety and enjoy!
Description
16Dim 8Alpha CAME
epsilon(eps)-pred Model
FAQ
Comments (6)
how many images did you use for your training data if you don't mind me asking. This looks really good. I'm new to making loras and only doing 15 epochs but looking at your results I feel like I should going more :).
The number of images I use for machine learning ranges from a minimum of around 20 to a maximum of around 100.
The number of repetitions is 1, and the number of steps is learned by increasing the number of epochs.
Number of batches: 2 to 3. 16Dim 8Alpha.
I use this PC as a heating source during the winter months, so I study Lora to coincide with my sleeping hours.(6~7hours)
I'm sure there are many ways to learn fast, so search around.
I believe there are settings that can be used to learn nicely for as little as an hour.
I use Affinity Photo2 image processing software to crop the background and process the image.
If you are interested in image processing I would recommend purchasing it as it will be sold at a discount during the Black Friday season.
I am not sure if this will be helpful, but I would like to mention what I keep in mind when I create a data set.
(1) Expressions of joy, anger, sadness, and happy (including ahegao)
(2) Front, back, left and right images
(3) Accurate tagging (important)
(4) Processing of white backgrounds using Photoshop, etc. (to portray Checkpoint's own backgrounds as beautifully as possible for use)
(5) Images of the upper body, lower body, and full body (tag each one properly)
(6) Elimination of blurred images (use images that show the face as cleanly as possible)
(7) When creating a character's Lora, remove tags related to eye color, hair color, hair length, and hairstyle and consolidate them into trigger words. (Optional. You do not have to do this.)
(8)As much as possible, tags for the same element should be consolidated into one of the tags to be called. (e.g., ① "pauldrons", "shoulder armor", ② "detached sleeves", "long sleeves", ③ "gauntlets", "vambraces")
My personal favorite optimizers are AdamW8bit, Prodigy and CAME.
I use CAME optimizer frequently these days, as I am a fan of new things.
I think Civitai's training function is also useful there, so if you have extra Buzz, you can enjoy trying it out in your spare time.
Let's enjoy our own pace 🤗.
xmilaidoux, can you, one day, also update your Tentacle and Witches characters loras to Illustrious too?
Also thank you very much for these releases and now updates to Illustrious style. All your loras are great!
I would like to try things out and upload them once I determine that I have reached a level where everyone can easily use and enjoy them.
Great work on the dynamic lighting images in the new set. Am I correct to assume that you tried and not exactly succeeded to create "meditation under a waterfall" scene, as well? Altogether gives me an interesting insight into to Illustrious 0.1 model.
Can you do Miwako Hazuki from Warau Kangofu?