Tensor.Art
Create

Kombagavad Gita -Furry Bara Style

LORA
Reprint

Updated:

3.7 Vishvarupa
v3.6Cumulonimbus
V3.5 Twilight

V3.5 Twilight：

What's New?

  • Restored Functionality: Some features from older versions have been reinstated for improved usability. This has unavoidably increased the file size, but hopefully, it's still manageable(doge).

  • Greater Base Model Influence: The final image's foundational style and the impact of artist strings are now more heavily influenced by the base model you use. This gives you more control if your base model is well-trained.

  • Broader Optimizations: Compared to the last release, this version offers more noticeable optimizations in several areas, including:

    • Art style consistency

    • Body structure and anatomy

    • Character facial features

Important Usage Tip

Due to the training on specific character features, you might occasionally see a particular pattern appear.

  • To avoid diamond-shaped body markings/tattoos, please add the following to your negative prompt: diamond (shape)

Version Detail

NoobAI
About this version This new version features an optimized training set, resulting in a different base style compared to previous releases. I've included 2.5D style images in the training data, which has significantly improved the LoRA's material generalization. You'll find that textures for objects, skin, etc., will now adhere more closely to your base model's inherent style. The biggest change in this version is the introduction of 9 distinct trigger words. Each one is trained on a different concept and carries a unique style, texture, and effect. You are encouraged to mix, match, and adjust their weights to find the perfect blend for your creations.Here is a breakdown of each trigger word, based on the training data and my own experiments: * an@t0my: Trained on images with excellent anatomical detail. This is intended to help improve and correct anatomical errors in your generations. * bu2nF1ght: Trained on images with a red-heavy color scheme or combat themes. Recommended when creating images with these subjects. * chr0m@t1cAbe2r@t1on: Trained on images featuring chromatic aberration or pale/pastel color palettes. Recommended for achieving these specific visual effects or moods. * c0nc8pt: Trained on images with strong character designs, including cowboy shots, busts, and full-body portraits. Use this to create well-defined character concepts. * d@rkF@nt@sy: Trained on dark fantasy-themed images. Recommended for that genre. * F@nt@sy: Trained on images with a distinct fantasy or hand-drawn/painterly scene style. Great for creating fantasy environments. m18ed: Trained on a diverse mix of themes, with a significant portion being suggestive or *NSFW**. Styles range from flat color to 2.5D. * ⚠️ Important Note: A large number of images featuring the character 'Sadiend' were included in this set, so this trigger may cause some overfitting towards her appearance. Use with caution if you do not want to generate images of this character. * P@le2e: Trained on images with a strong concept art aesthetic. Recommended for creating concept art-style pieces. * ske2chL1ne@rt: Trained on line art, sketch-style, or images with strong, defined lines. Recommended for achieving a line art or sketch-like effect. You can get a general idea of the styles from the comparison grids below, but I highly recommend experimenting to see how they work with your own prompts and models.

Project Permissions

Model reprinted from : https://civitai.com/models/1604504?modelVersionId=2156564

Reprinted models are for communication and learning purposes only, not for commercial use. Original authors can contact us to transfer the models through our Discord channel --- #claim-models.

Related Posts