Try size: 1024 x 1024 or 768 x 1024 or 768 x 1280Weights: 0.5~1.0Note: My top recommendation is this version, as it accurately reproduces the details and colors. / 我最建議的版本，細節跟顏色比較能還原。(Anima)Note: Z-Image Turbo successfully trained Rem, but failed to fully train Emilia, so I only did a rough training attempt. Those interested in well-known anime characters should research it themselves or download the dedicated character Lora. Z-Image Turbo already has a solid training foundation in "Doujinshi convention," so lower weights will be more flexible, while higher weights will be more stable. Regarding prompting, either natural language or tags are acceptable, so just use whichever you prefer. / z-image turbo有訓練成功Rem，但未能完整訓練 Emilia，所以我也只大概訓練一下。對知名動漫角色有需求者，請自行去研究，或下載專用的角色lora。原本z-image turbo就已經有很扎實訓練過"doujinshi convention"，所以權重更低會更靈活，愈高則會穩定。關於提詞，自然語言或標籤都是可以的，所以按照自己的習慣就好。(ZIDT)Note: I can't improve this overexposure effect, but I'll still update to v2. v1 was just too ridiculous; v2 is more stable. / 這種背景大量刷白的現象，我是沒有辦法改善的，不過我還是更新至v2。因為v1實在太搞笑了，v2會比較穩重些。(ILXL v2.0)Note: I don't recommend this version. I've already trained Illustrious many times... and I don't plan to try it again. A cfg value of 4.5 is recommended. Too low a value may result in insufficient saturation, while too high a value may result in excessive saturation. / 我不推薦這個版本，illustrious我已經訓練過很多次了...不打算再嘗試了。cfg建議4.5，太低可能飽和度不足，太高可能飽和度過高。(ILXL v1.0)

doujinshi convention booth / comic market / comiket

Files with SHA256: dbb062f08ac1c4a764b4d7a7664134ffb817f37ec10430557643003741dba723