Major change: From now on, each version will be responsible for a distinct manhwa style.
I Have To Sleep With A Stranger Illustrious + NAI:
How to use:
No trigger word for style.
Recommend weight from 0.8-1.3.
Due to low quality dataset, Controlnet and Upscaling are recommended.
Use "motion lines" if you want the lines to indicate action in the image.
Character name is supported, and is pretty close to the manhwa:
Elia, long hair, blonde hair, blue eyes, bright pupils, pointy ears
Elia, long hair, blonde hair, white dress, veil, habit, gold trim, covered eyes, detached sleeves
Ha-Eun, long hair, blonde hair, green eyes, large breasts
Ji-Ah, black hair, long hair, orange eyes, large breasts
Se-Yeong, medium hair, black hair, red eyes, large breasts
Sun-Yeong, long hair, black hair, red eyes, eyes visible through hair, large breasts, mole under eye
For Anima Preview 2:
Work with Tag and Natural Language caption.
Use @ihavetosleepwithastranger as the artist.
Add monochrome in the negative prompt if needed.
Another World MILF Hunter:
How to use:
No trigger word.
Recommend weight from 0.7-1.3.
Due to low quality dataset, Controlnet and Upscaling are recommended.
Use "motion lines" if you want the lines to indicate action in the image.
Character name is supported, but will not be close to the manhwa.
ILLUSTRIOUS (v5.0):
How to use:
Use the general trigger words "sylmhwa style".
"korean goddess" for a more beautiful face.
"realistic skin texture" for shinier skin.
"motion" if you want the lines indicating action in the image.
PONY (v4.0):
How to use:
Use any trigger words.
Use some pretty weird aspect ratios, but mostly 1:1, 3:4, 4:3, 2:3, 3:2, 16:9, 9:16, 21:9, 9:21.
FLUX (v1.0-3.0):
This is a Flux Lora for my favorite Manhwa style!
Trained locally using Flux 1dev Fp8 with T5XXl Fp16.
Prompting: I divided a prompt into 3 clauses in my training.
Clause 1 (Style): An illustration in the style of SylMS.
Clause 2 (The characteristic of the subject):
A woman with blonde hair posing.
A woman with long blonde hair and a white dress.
Clause 3 (The background):
The background is a church with stained glass.
Prompting Experience:
If the hand is ugly, lower the strength. The Lora needs to be fine-tuned, but fine-tuning does not support Fp8 yet.
If the art is in a digital art or painting style, add "anime" to your prompt. Flux Lora is experimental at the moment!
Refrain from using very long prompts because they will lead to a realistic image.
This model is NSFW, so please add SFW if you don't want a nude character.
Recommended strength is 0.7
My workflow (Drag or import to ComfyUI):
Description
v3.0 Update!
Clip skip 1 or 2. ddim for something strange!