This LoRA was made to address the issue of mainly sitting poses--at least for Victorian Anime Art XL and Noisy_Element_XL, since those checkpoints didn't have a substantial amount of sitting poses in their datasets. Knowing what I know about making checkpoints and LoRAs, I shall also make an update to those checkpoints.The dataset for this LoRA initially had 1474 images of various poses but I had to reduce it to 1000 for Civitai. As a result, some poses aren't quite as good as I'd like, so a different approach will be warranted later down the line. For now, this model seems to excel at sitting in chairs with/without legs crossed.What this model was initially designed to do:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0b1b6901-3131-4c38-b4fc-7cb53e256a1a/width=525/0b1b6901-3131-4c38-b4fc-7cb53e256a1a.jpeg" />I didn't want the word "sitting" to interfere with any of the other poses by becoming a trigger word, so instead I used the word "seated". If it was trained correctly, it shall have your character sitting on a chair or something.Since I had to remove many images and categories to fit into training for Civitai, there are only 13 trigger words for this first model. I've not tested them interacting with each other, but I highly recommend only using 1 trigger at a time, unless you know what you're doing lolA strength of 1 seems to work just fine on Victorian Anime Art XL and Noisy_Element_XL, but you're free to choose whichever weight value you like.Hopefully down the line (if I get enough buzz lol) I'll make revisions to this model and add more poses and stuff to it.Have fun~STFU (Set That Figure Up? As if...)This model is essentially the full version of STFD; all categories included, minus 400+ images to meet the 1000 image threshold for Civitai LoRA training.STFD_PI finally joined the dark side and have started making Pony-based things and have been pleasantly surprised. STFD may not make too much sense for Pony, as the overall pose quality is excellent. Perhaps STFD is too ambitious, as I had to sacrifice a lot of images to meet the margins, and sacrifice repeats in order to train it for less buzz. I believe it shows in the results, how it seems to struggle with such concepts as the high back kick, high kick, standing split, and on occasion sitting crossed legged. Perhaps I ought to break it down and make a more dedicated LoRA set.That being said, it seems to be good at the following poses:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5a526e20-d102-4756-9048-3069323d3247/width=525/5a526e20-d102-4756-9048-3069323d3247.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/415075e7-71bb-4db6-bd4b-4efdc7e6b6e1/width=525/415075e7-71bb-4db6-bd4b-4efdc7e6b6e1.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ae358013-54be-4355-9e76-3a6e40e5fa8d/width=525/ae358013-54be-4355-9e76-3a6e40e5fa8d.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/244f34d3-a07b-498f-9e6e-c92d3ef2d96c/width=525/244f34d3-a07b-498f-9e6e-c92d3ef2d96c.jpeg" />

STFD

Set That Foot Down... was the original name of it, but the scope of the LoRA now exceeds trying to simply correct sitting poses. This LoRA thus has many triggers. It's recommended to only use 1 trigger at a time, unless you know what you're doing.Note:The phrases "sitting on ground", and "lying down" are NOT trigger words; they're only there to supplement the actual trigger words "seiza", "wariza", "folded legs", and "on stomach"The word "Indian style" is like a co-trigger for "folded legs" in that, although it wasn't the trigger word for the dataset folder, that specific tag was used as much as the trigger word, therefore using them together might yield better results than if "folded legs" was by itself.

Files with SHA256: 20b6374f3842db9970f39f0d9b0502c91945571e9afaa09d88eaace680e90de3