ishin_Glittervip_style_FLUX

详情

模型描述

GlitterVip style LoRa Flux 1.D

Trigger words: glittervip_style

Prompt Examples:

glittervip_style, <subject> in glitter style

glittervip_style , (<subject> :1.3) , masterpiece, best quality, very aesthetic, absurdres

!! (LoRa is not very flexible, so you need to increase the weight of your subject hints as in the example) !!

All preview image generated with this prompts.

Weights: (0.7-1.2)

**Tested on:

ComfyUI + (model) Flux1-dev.sft

Forge + (model) Flux-dev-FP8**


Training Settings:

Kohya_SS (Local Training)

Dataset = 10 image

Epoch = 16

Training Steps = 1600

Optimizer = AdamW8bit

LR = 8e-4

此模型生成的图像