v2
I optimized the training dataset and parameters. This version has higher stability and better generation quality compared to v1. Will remove if requested.
Trigger Words:
t1f style, monochrome, Sketch art, line art, drawn,Other Data:
step:30 / cfg:7 / Sampler:EULER A / Schedule type:Karras / checkpoint:WAI-illustrious-SDXL_v17 / Strength recommendation:<lora:TTF_Style_v2_R11:0.9>________________________________________________________________________________________________
v1
Using TTF's artwork as a training set.This artist deleted all his accounts at 2026.2. But I don't know what happened?If you know anything about this artist, please share in the comments! It's such a pity that such a great work has vanished .so I training this style lora.Will remove if requested.
The photo editing tool comfyui
step: 30
cfg: 7
Sampler: EULER A
Schedule type: Karras
checkpoint:
WAI-illustrious-SDXL_v17.0
Strength recommendation:
<lora:TTF Style_R15:0.7>
Trigger Words：
t1f style
(P.S.: Image quality is relatively balanced at a model strength of 0.7. Increasing the strength will result in stronger image tension, but the images will also be very prone to structural breakdown. The only exception is when the character being trained is Tatsumaki; setting the model strength to 1.0 does not cause structural breakdown and the image tension is excellent (for example, image 4 has a model strength of 1.0). This is because Tatsumaki was the most frequently used data in training this LoRa dataset. If the feedback on this LoRa dataset is positive, I will try to train a more stable version when I have time.)
Description
FAQ
Comments (8)
很高兴还有人记得这个作者。25年这个作者就停更了，在外网一些论坛上这个作者被骂的很凶，说他作品里画的丑男是带入他自己本人，诋毁作者的外貌。不知道是不是因为网暴的原因，也有可能是作者找到正经工作要跟以前划清界限了。总之这个作者确实可惜了，很喜欢他的画风和叙事风格。
hi,can you make exodia girl's style?
Okay, I'll try it when I have some free time in the next few days.
@Troen Thank you
@View_ Hey bro, the artist's published work was created using AI. The images have many issues with perspective and lighting on the human body, as well as numerous anatomy errors like knuckles. Since it's an AI-generated image, I don't think using it as LoRa training material will yield good results. It might even lead to the generated model producing more errors and messy images.Please reconsider whether you still insist that I use this artist's image to train the LoRa model.
@Troen Oh, I didn't think of that. If you think it's not feasible, then forget it.