NNGMIX-XL版本v3.0 (ID: 1194831)

start quality prompts

masterpiece, best quality, amazing quality, very aesthetic, newest, general, 

negative prompts

lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract], multiple views, censored, bar censor, mosaic censoring, red pupils, colored inner hair, 

v3.0

With the release of NoobAI vPred 1.0, I’ve performed another merge while maintaining the TE.

In this version, I’ve included a small portion of the WAI-NSFW-illustrious-SDXL model.
The usage remains the same as the previous version.

v2.0

This version was created by merging NoobAI vPred 0.9R, similar to N40NAilousV2-xl.
The TE of the vPred version has been preserved, and the final touches were made by partially merging it with NTRMIX XIII while adjusting the UNet.

The usage is the same as the previous version.

v1.0

This is a newly created merged model based on NoobAI 1.1, departing from the previous NTRMIX-based merging approach.

The model was merged with the goal of enhancing quality by utilizing Quality LoRA. At the final stage, the OUT blocks (1 to 4) of N40NAillousV2-xl were adjusted to slightly influence the color tone and output results.

Considering the potential use of LoRA, the block weight values for LoRA were carefully adjusted during the merging process, so there should be no issues when using LoRA.

The model is named NNGMIX-XL, deriving from the name NGMIX, which was used in SD 1.5. The "N" represents either "New" or "NoobAI," reflecting its origin.

I generally believe that it's better for users to freely explore without being bound by specific prompts or settings. However, as an example, I’ll share my preferred usage method.

Thank you once again to those who asked about the usage in the comments.

• Quality tags, Rating

• Character

• Clothing, Actions

• Background, Surrounding objects, Lighting, etc.

• Artist, LoRA

Settings:

• Sampling: Euler a

• Schedule: Automatic

• Steps: 28

• CFG Scale: 5

• Width x Height: 832~1216

When using Hires. fix:

• Upscaler: 4x-UltraSharp

• Steps: 10

• Denoising strength: 0.35

• Upscale by: 1.25

ADetailer: Only used for faces

The above sentence was translated using ChatGPT.

描述:

训练词语: