King Azwel アズウェル王 | Hollow Earth Isekai 空洞地球の異世界 劍魂VI (SOULCALIBUR VI)
Version 1 : From 25 P
Can you support me 500 Buzz or more ^_^''
(make any one of these Characters Lora = 1000 Buzz , And my Approve)
(1) Directly posting your images 🖼️ to the model's page will give me Blue Buzz 🔵⚡
(2) Gives me Lover ❤️ badge
(3) Gives me Yellow Buzz⚡!
Lora : 0.5 - 1.0
Prompts :
score9, score8up, score7up, score6_up,
wooden room, indoors, sitting on Chair, Sitting at attention, From Front,
1boy,solo, 18-year-old,
King Azwel Hollow, KingAzwelHollow, King_Azwel_Hollow,
a man in a white suit,crown on his head,red and gold crown on his head,
Black hair, medium hair, glasses, cape, crown, velvet crowe,
white gloves, gold trim, armor, faulds, gauntlets, shoulder armor,
white and black outfit,
detailed decoration,parted lips,looking_at_viewer,moody,
expressionless,
masterpiece, best quality, amazing quality, very aesthetic, newest,
(anime_screencap:1.3), (anime_coloring:1.3), official style, perfect quality,
N:
score_5, score_4, score_3_down, ugly, poorly drawn face, blurry, blurry face, (3d), realistic, muscular, blurry eyes, poorly drawn eyes, patreon, artist name,lowres, worst quality, bad quality, sketch, error, bad hands, logo, text, watermark, username, date, ugly, distorted, censored, multiple views, 2koma, 4koma, comic, manga, bad anatomy, bad proportions, extra limbs, extra digit, extra legs, extra legs and arms, disfigured, missing arms, fused fingers, missing fingers, unclear eyes, mammal, furry, ambiguous form, feral, semi-anthro, 2 girls,
Checkpoint: 熟メス儀式 | Mature ritual (Illustrious)
https://civarchive.com/models/994401?modelVersionId=1294629
Description
Comments (13)
Hello, I was looking through your training data and I was surprised to see how simple and "low-quality" your images are. I recently started learning how to train my own LoRAs and I feel like you are breaking all the rules here.
Is there a method to your madness? Are you not limiting the potential of your LoRA by training it this way? Or does that make it more versatile?
Thanks for any feedback.
As someone said : more images in many angles always it will be better
https://civitai.com/articles/2833/need-150p-for-this-reason
I learn real things in this Lora, Even in SD1.5 , AI can understand 3D
So you need to crop many angle for each outfits,
【KK_REAL】Joker Heath Ledger | Dark Knight
https://civitai.com/models/66034/kkrealjoker-heath-ledger-or-dark-knight
You can download Training Data and check it
Even in SD1.5 , Only 20 pics can make a lora
That's mean, in iLL (Illustrious) You dont need that much pics
Honami Okudera 奥寺穂奈美 | Saimin Seishidou 催眠性指導
https://civitai.com/models/362143/honami-okudera-or-saimin-seishidou
In iLL , You only need to found out , which checkpoint is the best.
Checkpoint: 熟メス儀式 | Mature ritual (Illustrious)
https://civitai.com/models/994401?modelVersionId=1294629
And I choose this
I train iLL Lora in https://civitai.com/models/train
Because it is perfect setting and you dont need change anythings
( I only change the network Dim to 4 = 27MB)
All my loras is good and use minimum bytes
I tried Network dim to 1 = 7MB
but not good
I dont understand why people make 218MB Lora
and still download them
when someone train a Lora in 218MB, that's mean he dont really understand AI
You need more angle and now this LORA
King Azwel アズウェル王 is my character
I make in with https://gemini.google.com/u/3/
That is a new technique and new skill about
How to use gemini
King Azwel アズウェル王 only 25 pics
but you still need more angle
that is the point
masterpiece, best quality, amazing quality, very aesthetic, newest,
(anime_screencap:1.3), (anime_coloring:1.3), official style, perfect quality,
This prompt is useful
I really concern about "Scale" Face
So many people Do Not concern about "Scale"
My Lora comes out a good "Scale" Face and body
But that is concern about "milfs" or "kids"
I dont like kids face, that is ugly.
SO I only use this checkpoint and this version
Checkpoint: 熟メス儀式 | Mature ritual (Illustrious)
https://civitai.com/models/994401?modelVersionId=1294629
Why Train lora with
( I only change the network Dim to 4 = 27MB)
All my loras is good and use minimum bytes
I tried Network dim to 1 = 7MB
Because Computer language =
1, 4, 8, 16, 32, 64, 128 .....
So I only can choose
1 or 2 or 3 or 4
the only answer is 4 , Dim to 4 = 27MB, this is the perfect size about lora
If you need to train best quality
You can use big size pics, and much more pics
and dont change the Network dim
come out also 218MB
But that wasted my time, I dont want to "work" on "train Lora"
I dont want "train Lora" = my job
that is wasted my time
Use this
Bulk Image Resizing Made Easy
https://www.birme.net/?target_width=1024&target_height=1024
In fact : King Azwel アズウェル王 OUTFIT = soul calibur 6
I make in with https://gemini.google.com/u/3/
That is a new technique and new skill about
How to use gemini
gemini Prompt :
2D CHARACTER SHEET
Generated 2D character angle color setting background pure white
Inclusive Front side Inclined side Back left side Right side
將以上人物 變成一張 2D CHARACTER SHEET 生成2D人物角色設定圖 背景純白色
包括 正正面 傾斜面 背面 左面 右面
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.