Lora of Guardian tales chibi style

[English] (google translate)

The lora work well between 0.5 and 0.8, but i recomend to set it to 0.7 or below.

if someone asks why I put 3 images of robin, it's because I'm a simp and I liked all 3 images.

Recommendations

• Use checkpoint where it is easy to create chibi character, like akkaimix or others similar to that one

• If you use a checkpoint where it is difficult to make chibi characters use (chibi:1.3) in the prompt (you can also use the tag "full body", but using that tag will mess up the background sometimes).

• This model works better if you use it together with another character lora. (this is highly recommended, does that fix a lot of things)

Non-useful Info.

I created this model in March 2023, this was my first attempt to create Loras of styles. Originally I had planned to upload the original version which weighed 144MB but in the end I decided to re-train the model using the same dataset and reduce the weight.

The dataset was automatically tagged, at that time I didn't know very well how it worked so I left everything as is, now that I know a little more how to tag I will probably revise the dataset and fix some problems, and then upload an updated version.

[Chilean] (native chilean)

Este fue mi primer intento de crear un lora que sea de estilos, en resumen la wea me quedo como el pico pero al menos funciona relativamente bien con otros loras de personajes y weas random, como la wea la hice el año de la re corneta voy a revisar el dataset de la wea y arreglar todas las pifias de la caga porque la wea la hice a puro autotagging y esa wea funciona como el pico, Entonces luego de fixear todas las pifias subire una version hd 1080p 4k y toda la wea esa

PD: pal ctmre enfermo que se atreva a mencionar que esta caga esta escrita como el pico le informo que pase lenguaje con un 4.0 y que tiene toda la razón