OK (AI image)🔞🏳️‍🌈

AI製圖 (AI-generated image)
SD module：Lawlas's Yiffymix 2.0 or Crosskemono2.0
中文 O / English △
只是個IT人員在下班時學習新技術
furry,gay,獸人,ケモノ,兽人,男性人外,muscle,bara furry,NSFW

[OK (AI image)🔞🏳️‍🌈 @xd123437]

statements
➊ Just an AI user.
➋ Only use public MOD and lora, etc.
➌ Private MODs are trained for learning purposes and will be written as “訓練” or “Training”, etc.
➍ Still learning. Sorry for the glitchy image.
➎ Posts mainly for learning, experience, and using AI.

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@nike857857 https://t.co/NvZBfa9qYD

[OK (AI image)🔞🏳️‍🌈 @xd123437]

[AI]藍龍
ControlNet1.1 inpaint與Tile 用起來還行
模組為crosskemono2.5為底A混nijiv5style(換色練習有連結)各50%
接下來不生圖了，除非SD有特殊差件或重大更新，去研究其他AI
won't generate more images unless SD releases specific components or significant updates
look into other AI tools

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@ArcticPlatinum You could also try NijiJourney V5, which is much simpler than SD and feels more like using a
search engine, making it very user-friendly. However, it is a subscription-based service that
requires payment.

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@ArcticPlatinum It's like writing code where you use downloaded modules and optimize poorly performing parts
with plugins or rewrites Prompt. Each generation is a try-and-error process. AI generation often
involves tweaking details, whether it's through AI inpainting or manual redrawing.

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@ArcticPlatinum because it's kind of like programming

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@ArcticPlatinum (muscular male (adult) furry),white fur,fox furry,long hair,Japanese Kimono,(open clothing),(close_eyes:1.5),long_sleeve_Kimono,long_wavy_hair,Highly detailed,indoor,sky,lantern,fluff Tail,(soft fluff fur:1.2),8K,hand of guido daniele,perfect lighting,complex background,abs

[OK (AI image)🔞🏳️‍🌈 @xd123437]

[AI]換色練習
步驟1：先用Crosskemono模組生成(圖1)
步驟2：用ControlNet的HED提取線條與控制步驟3生圖
步驟3：換nijiv5style模組進行文生圖並使用cutoff與提詞控制顏色影響(圖2)
ControlNet
https://t.co/284AvWKQpH
NijiV5style
https://t.co/bI8E49XmMr
cutoff
https://t.co/NpsDJnEtgS

[OK (AI image)🔞🏳️‍🌈 @xd123437]

最新適用於ControlNet的poss調整器， 可以隨機生成、圖片偵測、手動調整動作與手勢
作者架的網站：
https://t.co/AGu5tPeuRH
作者開源碼：(如果你想自己架站或網路不穩定) https://t.co/EuCupJdYvw
中文教學網站：
https://t.co/S3COAkKUhH

[OK (AI image)🔞🏳️‍🌈 @xd123437]

目前測試附加功能圖像放大使用4x-UltraSharp為主與SwinIR_4X為副要演算法，演算法可見度設定為0.5~1的放大效果最好，4x-UltraSharp點擊下方連結把pth檔案下載後放到models\ESRGAN底下再重開就可以使用了(有人試出更好的放大演算法歡迎交流
https://t.co/B32WhudOb2

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@CoheriV SD has a checkpoint containing NSFW content and you want to test it by inputting simple prompts like 'p●nis', 'cum', 'sex', but if you want to use a checkpoint without NSFW, you need to use TI and LORA to teach AI about the concept of NSFW, and LORA need trigger words.

[OK (AI image)🔞🏳️‍🌈 @xd123437]

[AI] 口爆 blowjob
我感覺我不太會使用AI...
想試試nijijourney v5，但它不能生成色圖...
I feel like I won't be able to use AI...
I want to try NijiJourney V5, but it doesn't allow the generation of NSFW images...

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@wolfcharizard 基本上我只剩下換顏色與骨架模組還沒有使用過，等這兩個都使用過後除非有重大技術更新，或換到使用SD的公司當IT，不然我應該不會再使用AI生圖了，預計就玩到這個月底，畢竟細膩的筆觸與細節是AI目前無法模仿的，除非生物電腦有出現重大突破或出現強人工智慧

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@KennyWuKai https://t.co/3tKq3GNJ2V
這邊有大概寫怎樣可以唱出比較正常的，大致流程，可以的話先別唱色圖，SD的AI大家都是從正面照片先開始練的(大概？

[焰虎_Tako @banna097]

@xd123437 有裝EasyNagetive 模組是用Crosskemono2.0,lora和VEAa還沒碰到
算出來蠻多破圖的

[OK (AI image)🔞🏳️‍🌈 @xd123437]

[AI]腳腳與肉球
MOD：Crosskemono2.0
lora：purefurry

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@KennyWuKai hypernetwork 很難用，現在用lora比較多，也比較好用，我的作品也不算很好跟nijijourney v5收訂閱費Discord的AI圖真的差太多了，我還在學習怎麼用開源的模組調整的更好

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@Putin0317 謝謝❤

[OK (AI image)🔞🏳️‍🌈 @xd123437]

英文應該是 I feel like I'm not familiar enough with AI... 才對...
我怎麼會用be able to...

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@banna097 你可以用小畫家塗成你想要的顏色再用【圖生圖】進行圖片重繪，重繪權重你可以需要慢慢調，當然也可以重繪第一版後再用��繪出來的圖進行AI的2次重繪，或是用【局部重繪草稿】來上色，或優化提詞想辦法告訴AI腹部不要是白色的

[OK (AI image)🔞🏳️‍🌈 @xd123437]

[AI]再一發魔法陣就完成了
跟前一張一樣的mod&lora
優化提詞讓顏色可以更加鮮明
下體會不會太大(?

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@banna097 https://t.co/FZ8YWvX2fz
https://t.co/KMk8CrGL9W
https://t.co/RM9s2LbQJw

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@banna097 在civitai用新模組時，可以點開他模組展示圖最好的圖片，他都會寫他怎麼設定可以創造出該圖，先依照各模組示範圖設定進行生成，再加入自己的想法慢慢修改提示詞

[OK (AI image)🔞🏳️‍🌈 @xd123437]

@banna097 通常畫獸人都會再下載boring_e621與bad-image這兩個TI，bad-image版本很多你再慢慢試��個比較適合你，因為每個人Commit hash都不同所以我也不知道哪個比較適合你，反向提詞TI，放在反向提示詞的方框