Nov 9, 2026

【公告】關於近期生成速度變慢及運算資源緊張的說明 — 親愛的 PixAI 社群成員：感謝您一直以來對 PixAI 的喜愛與支持！近期 PixAI 的生成速度較平時慢，排隊時間也比平常更長，造成不便，我們深感抱歉。原因說明在 Tsubaki.3 上線後，大家的使用熱情帶動 GPU 資源需…

查看詳情

SDXL

MixNoobMerge_V1-Vpred

#獸人

上傳時間: 2025年6月7日下午2:06
使用數: 250
評論: 非常好 (2)
取樣步數: 20
取樣方式: Euler
CFG Scale: 3
反向提示詞: bad quality, worst quality, worst detail, low quality, low res, bad anatomy, mutated, deformed, fused, conjoined, extra tail, bad hands, bad eyes, ugly, jpeg artifacts, text, watermark, artist name, username, logo, signature, patreon,
許可權: 允許圖片生成與分享
允許使用者下載此模型
允許生成圖片用於商業用途

描述

https://civitai.com/models/1652619?modelVersionId=1870588

Remember to set the prediction type to v_prediction so Vpred models generate the images correctly instead of artifacts.I recommend using High Priority(site)/Rapid Generation(app mobile) on pixai.art to generate the image faster, as in normal mode it can take 2 days to generate the images.

V1-VPred:

Model created from merges with the models:

similarity_add_difference, Alpha 0.95, Beta 0.05

merge + MixIllustriousMerge6

similarity_add_difference, Alpha 0.7, Beta 0.3

DeepViolet VPR + NAI V-Pred + NoobAI_VPR_IV_4