I was thinking about whether it would be better to not train this since the limitations of SD 1.5... The images have too many details, don't think SD1.5 will produce a good result...

But ...why don't give it a try?