Gloria Sol v2 XL

V2 was trained with JoyTag (JoyCaption) since manual tagging of 770 images is way too time consuming. I have included a single tag to differentiate the zoom levels. The rest of the tags /description is from JoyTag. In the prompts you can call her gigi, which is also a tag.

You need adetailer or inpaint to fix the face if you generate full body images or if the LoRA weight is set too high.

Some of the preview images are using the DMD2 LoRA which is also helping to get better results.

Model 1: Big Lust v1.0 (SDXL) Brighter images (I have used this for training)

Model 2: Big Lust v1.5 (SDXL) Darker images with minor quality loss

SDXL ≠ Pony You should use SDXL models for generation if you want realistic results.

Training Resolution

Resolution 1: 832x1248, 2:3, Portrait
Resolution 2: 1024x1024, 1:1, Square
Resolution 3: 1248x832, 1:1, Landscape

Training time v2: 28.25h (RTX 4090)

Example prompts

g1g1_face, close up photo of gigi face ...

g1g1_portrait, close up portrait of gigi wearing a suit ... 

g1g1_body, photograph of gigi, gigi walking in a forest ...

Gloria Sol v1

Requested by Caius

Training Data

This LoRA was trained with images of her face, upper body and full body. It will create NSFW content if you don't prompt for clothes and add negative tags.

model: epicphotogasm_lastUnicorn

Image count: 736

Resolution 1: 627x941, 2:3, Portrait, 70%

Resolution 2: 768x768, 1:1, Square, 30%

Example prompts

g1g1, close up photo of a woman showing her face

g1g1, photo of a woman showing her body

g1g1, portrait photo of a woman

描述:

first release

训练词语: g1g1

名称: GloriaSol_v1.safetensors

大小 (KB): 295027

类型: Model

Pickle 扫描结果: Success

Pickle 扫描信息: No Pickle imports

病毒扫描结果: Success

Gloria Sol

资源下载

下载价格VIP专享

仅限VIP下载升级VIP

犹豫不决让我们错失一次又一次机会！！！

原文链接：https://1111down.com/1015018.html，转载请注明出处

Gloria Sol版本v1 (ID: 256515)