Alive:The Final Evolution -最終進化的少年- doujin ver 1.0

LORA
Original


Updated:

“オレの居場所は……!”

“我的立身之地…”

“我的容身之處…”

“I belong with…!”

 

No shotacons could read this part without feeling deeply shaken. Standing on the brink of life and death, Yuuta had only one thought—Taisuke. And this line was his truest confession.

 

I’ve reread just those last few pages of Volume 3 more times than I can count. It's etched themselves into my memory, leaving behind a sense of peace that feels almost detached from the world. Every time I revisit them, I’m reminded of that stillness.

 

May the author, and every reader—those still living, and those who’ve passed—forever find that same quiet peace.

 

Next comes the usual nonsense time. This was my first time trying to make a LoRA using only black-and-white information. In fact, the color details of the characters are quite vague. Yuuta is fine—basically all the colored illustrations show him with brown hair and brown eyes. But Taisuke is different—every colored image varies drastically, and I had absolutely no idea what his hair or eye color was supposed to be. In the end, I chose black hair and purple eyes, which appeared more frequently and also matched the general impression from fan works.

Situations where the “stereotypical image” differs greatly from the original always come up. Yuuta’s design even changes from panel to panel, making it quite difficult to reproduce accurately. And as for current mainstream training models like Illustrious 2.0, they tend to learn certain neutral or even feminine traits for young boys—something I’m sure many people have noticed across various LoRAs.

As for Taisuke, in the original work his hair is actually quite soft and smooth most of the time—not the spiky, triangular shape seen here. But that spiky version is the one most widely accepted in fan works, and it’s also the style AI picks up naturally, so I just kept it as is.

I tried many times, but if the training set was made entirely with AI-redrawn images, Yuuta’s likeness became way too inaccurate, and far too feminized. In the end, I had no choice but to include black-and-white material in the training set, so you might encounter some weird bugs when using it—try a few times.

 

As mentioned above, Yuuta’s likeness is much harder to reproduce compared to Taisuke. But I’m glad that some of his more iconic expressions—like “grin” and “smirk”—were at least somewhat faithfully preserved. So, here are the trigger words:

 

瀧沢勇太: atuuyawazikat,yuta-very short-brown hair, yuta-brown eyes,      

 

叶太輔:   uonakekusiat, taisuke-black hair, taisuke-darkpink eyes,   

 

Expressions: "grin", "light smile, closed mouth", "smirk:0.1"

 

Based on initial testing, there aren't any specific negative prompts that need to be added. Only in very rare cases does Yuuta's hair appear too rounded—in those cases, you can try adding (((pixie cut))), (((bowl cut))). In most situations, it's not necessary and makes little difference.

 

In the vast majority of cases, Yuuta and Taisuke should maintain a relatively stable body size difference. This has already been reproduced as faithfully as possible. If the body sizes appear too similar, you can add "height difference, age difference, onii-shota"—use them together, separately, or with adjusted weighting as needed. Probably due to how the dataset was designed to emphasize the size difference, generated images always show Yuuta in front and Taisuke in the back. There's currently no effective prompt to adjust this layout.

 

Only the early-series designs are included—none of the time-skip designs from two years later—so please don't attempt to generate those.

 

CFG: 3 – 5

 

Model Weight: 0.7 – 1

 

Sampler + Scheduler: dpmpp_2s_a + beta (or adjust according to checkpoint)

 

30 steps seems to work best.

 

Resolution: no specific requirement.

 

In this era where AI-generated images are still plagued by controversy and criticism—over-commercialized, fast-food-like, lacking creativity, emotionless—I understand. I agree, even. So much of it is merely about indulging fleeting desires.

But I also disagree. Because when AI is used to awaken dormant memories, to make forgotten moments shine again—no matter how briefly, like a falling star—that spark is something real I’ve offered to the world of AI. And I will continue to follow love.

 

The next LoRA is already in the works. In this restless age, most ACG works exist only to deliver quick, fleeting pleasure. But there was one anime that truly captured the confusion and vividness of boyhood—a little helplessness, yes, but also full of life. Sadly, such sincerity couldn’t withstand the weight of commercial pressure, and the second season was cut short. A bitter end that still leaves me unsettled.

Perhaps you already know which series I’m talking about. The schemes of adults couldn’t stop the boys from moving forward—they still press on, toward the future.

 

This will be a major project, planned to include at least seven characters. With my current setup, I can’t manage them all—so I’ll start with the ones I love the most. I’ll be disappearing for a while. Hopefully, by the time I return, Tensor will come back.

 

Until we meet again~

Version Detail

Illustrious

Project Permissions

Reprinting is strictly prohibited

    Use Permissions

  • Use in TENSOR Online

  • As a online training base model on TENSOR

  • Use without crediting me

  • Share merges of this model

  • Use different permissions on merges

    Commercial Use

  • Sell generated contents

  • Use on generation services

  • Sell this model or merges

Related Posts