bismillah
its really simple, u see as thanos once said
"use the stones to destroy the stones"
this gave me the idea on how to convert it between 1.5 and xl
step 1 gather a dataset by using the lora itself to make a dataset
basically u need to use the pony or sdxl lora to generate urself a dataset, use the lora itself to make it into a dataset, sorta like extracting the dataset but its unlimited since u can change the dataset however u want and since the quality is really good its reliable, if not then u could just prompt it to be reliable in the dataset
important thing to note is that pony uses less dataset images so u need to gather a large number of images from it to make it into 1.5
step 2 sorting and training the lora like normal
for sdxl/pony u need an average of 50 images more or less to get a good style
for 1.5 u need an average of 100 or more to get a good lora too
ofc both factors depend on quality and how complicated ur style or character or concept is
after u calculated that all u need to do now is follow my youtube tutorial (its not updated with pony but it can be used to add tags to ur dataset)
step 3 enjoy ur lora
thats it u reached the end, u can easily just train a pony lora on civit ai itself
remember to always credit the original creator of the other sd lora version
(theft is cool but not stating the original source is cringe)
if u want any lora convertions ill be happy to make it for u for absolutely free
just add me on discord through my username xmattar
but first u need to read my lora making article