Ever wanted to make a LoRA but didn’t know how? I’ve got a general purpose guide for that. However one thing I don’t have in that, and know the lack of an example for intimidates newcomers sufficiently, is a full tagged dataset. I’ve finally addressed this omission for characters! My new Sherlock Holmes LoRA was made entirely off of public domain depictions of the very public domain character, and as such no IP law prohibits me from sharing these pictures and their tags! To DL the datset, just click "Download" to the right of "Training Images". I do note that this a much lower quality dataset on average than one would prefer, monochrome magazine illustrations tend not to have the highest quality printing or preservation quality, but it works and is plenty sufficient to get the idea that you can translate to more modern characters.
I’ve wanted to do this kind of thing for a while and finally decided to do it. I’ve actually done a background off of PD and CC images before with my Angkor Wat LoRA, but the dataset is for a location/background concept, when I’m sure most new creators want characters, and was something I made relatively early in my LoRA making (so not entirely what I’d recommend now). Also it’s a lot larger than most LoRAs need due to trying to teach a broad concept than a relatively narrow one, and image quantity needed is something new creators often overestimate due to older guides (In my experience it’s 20-30ish for a single outfit character for Illustrious. The Holmes one is a bit bigger to ensure artifacts don’t get interpreted as part of the character and because I had the data handy).