Try size: 1024 x 1024 or 768 x1024 or 682 x 1024 or above
Weights: 1.0 (about 0.8~1.0)
Note: The rendering effect of this current LoRA is the closest to Doubao AI's composition. However, in terms of flexibility, ZIT remains the optimal solution, despite its higher error rate. (ZIDT)
Note: The overall realistic effect is exaggerated, after all, Anima was mainly trained for anime. (Anima10)
Note: This lora has a lot of problems with the girls's arms and legs, so It takes many times to adjust the prompt and lora weights. If you need front photo, please add "front view" and "facing viewer". (FlUX)
Note: Recommended ADetailer, if base model had baked vae, please closed Vae in ADetailer VAE. Detection model confidence threshold adjust "0.7", because there are too many people in the background. Some basic models are not suitable, so replace them if they are not suitable. (SDXL)
Description
FAQ
Comments (1)
Someone seems to have set the "Distilled CFG Scale" 3.5 to 1, and then the generated pictures kept getting errors, so they left negative reviews. This is a self mix-up incident, and is very suitable as a negative teaching material, so I won’t hide the picture. (The author has this permission) 有人好像把 Distilled CFG Scale : 3.5 設定成 1，然後生成圖片一直錯誤後，就留下差評了。搞烏龍事件，很適合當反面教材，所以我就不隱藏圖片了。(作者有這個權限)
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.