弄了三個月AI才解鎖新技能/Took three months for the AI to unlock a new skill (Patreon)

大家好!我是宮堡雞汀

今天7月12日，正好是開始練習AI滿三個月

直到昨天才開始使用ControlNet這功能

用了才發現，讓我想起哈利波特一個梗圖

榮恩!聽我說，這東西比魔杖還好用啊

以往設定多人都得用亂數特徵去抽

比方說A的特徵會跑到B身上，B的特徵會跑到A身上

就要賭那些特徵都剛好在原本的角色身上，否則就會變成不同人啊

用抽卡遊戲來說就是抽UR那機率的概念

-

現在用了ControlNet的功能後，整個難度驟降

大概從UR變成R卡或N卡那樣的輕鬆啊...

不過後續要修圖又是另外一回事了......

-

也多虧了這功能，好讓我的原創角色可以好好的聚在一起啦!

之前沒使用這功能光兩人好好出現就超費時

不過因技術受限，目前上限好像3人為限

為了不要有排擠人的嫌疑，只好編個理由讓純玲不在現場了

比如說跟誰去約會之類的w

將來用在H場景或許某些人就不用以無臉或大眾臉身分登場啦

-

總之為了這一張圖

幾百張的圖就這樣躺在了資源回收桶

往往花費的時間心力比同人角色還多

而且手指真的挺麻煩的，最花時間的地方

能藏就藏，而背景路人通常比較不重要

當跑不出你要的結果只好趕緊模糊省時省事

不然一堆圖還等著排隊要修啊...

-

雖然原創角人氣低是可以預料的

況且也不是H場景

不過為了豐富這些角色個性與生活

我認為這是必要的

-

相比同人角色多用敘事的方式描寫而不用對話

一來是怕會破壞原作角色人物的性格

二來是怕自己對角色不夠深入了解描寫得不夠好

但原創角色就沒這問題，只不過在日文語氣用法上還是要有所區分就是了

當圖做出來就是開始想角色之間的劇情跟對話了

通常發在PIXIV的日文翻譯，自己還會再看個幾遍看哪邊有沒有需要修改

也是非常花時間，不過辛苦也是值得的

感謝各位看到這邊

https://www.pixiv.net/users/591455

若喜歡我的作品不妨訂閱支持

===========================

  Hello everyone! I'm KPChiken. Today, July 12th, marks the beginning of my three-month journey practicing AI. It wasn't until yesterday that I started using the ControlNet feature. Once I tried it, it reminded me of a meme from Harry Potter." Ron! Listen to me, this thing is even better than a wand! "Previously, when setting up multiple characters, we had to rely on random features to assign them. For example, Character A's features might end up on Character B, and vice versa. We had to rely on the chance that those features would align with the original characters; otherwise, they would become completely different individuals. It was like the probability concept of drawing a UR card in a card game.

  Now, with the ControlNet feature, the difficulty level has significantly decreased. It went from drawing UR cards to a more relaxed R or N card-like experience... However, post-processing the images is a whole different story...

  Thanks to this feature, I can finally gather my original characters together! It used to take a lot of time and effort just to depict two characters properly. However, due to technical limitations, it seems that the current limit is three characters. To avoid any suspicion of excluding someone, I had to come up with a reason for Sumire not being present. Maybe she went on a date with someone, for example. In the future, when it comes to adult scenes, certain individuals won't have to appear as faceless or generic characters anymore.

  In the end, for the sake of this one image, hundreds of discarded images ended up in the recycling bin. Often, the time and effort spent on them exceeded that of fan-made characters. And fingers... they can be quite tricky. The most time-consuming part, really. If they can be hidden, they should be. Background pedestrians are usually less important. When unable to achieve the desired result, blurring becomes a quick and convenient solution. Otherwise, a queue of images awaits their turn for editing...

  Though it was expected that my original characters would have lower popularity, especially since this isn't an adult scene, I believe it is necessary to enrich their personalities and lives.

  Compared to fan-made characters, which are mostly described through narration rather than dialogue, I fear that it might disrupt the established characters' personalities. Additionally, I may not have a deep enough understanding of those characters to portray them effectively. However, with original characters, that isn't an issue. Still, I need to differentiate their usage of Japanese language and speech patterns. Once the images are created, I start thinking about the plot and dialogues between the characters. Usually, I post them on PIXIV with Japanese translations, and I even review them several times to see if any modifications are needed. It is time-consuming, but it's definitely worth it.

Thank you all for reading this far!

https://www.pixiv.net/users/591455

If you enjoy my work, consider subscribing to show your support.