I would prefer if you guys don't use this model for Commercial Use.

You can check out my pixiv where I use my lora to create images.

https://www.pixiv.net/en/users/34413372

training realistic lora for pony can be hard so I would recommend the cyberrealistic v6.5 lora or sdxl lora

ponyxl: v3 (cyberrealistic version) (includes A and B variant)

this model have over 600 images and the improvement comes from that the model can create male characters also, the v2.5 struggle with that. I recommend lora strength 0.7

This model is not only a style model, but also a concept model.

A and B variant model are trained on the same dataset, but I have trained both differently.

use this model. the lora only works with this. CyberRealistic Pony 6.5, https://civarchive.com/models/443821?modelVersionId=953264

ponyxl:2.5 (cyberrealistic version)

use this model. the lora only works with this. CyberRealistic Pony 6.5, https://civarchive.com/models/443821?modelVersionId=953264

I reduced the dataset from 600 to 200 and improved the caption. I have manually added the caption for every image. use 1woman or 2woman's instead of 1 or 2girls, same for boys. I recommend you use the prompt that I have used on the images that I uploaded here on civitai. this lora need specific keyword to work properly.

go with lora strength 1/weight 1

pony xl v2:

the dataset is the same I used on flux dall e 3 model with the same caption and because of that you can somewhat use natural language. You can still use the tags, but it won't be effective. Use the prompt in images as an example I showed to get better quality

sdxl:

I used over 600 images and I recommend using Juggernaut XL as checkpoint

go lower than 6.5 cfg for better result

pony xl:

I used Dall-E 3 to generate the dataset and used around 50 images of them. This lora is made to create the Dall-E 3 woman only. add 3d in negative for better images