Preview images use only Pony Diffusion V6 as a checkpoint and no other LoRAs.
❌💲Commercial use of this model is not allowed.
I trained this LoRA on my own art only, for this reason it has stricter permissions than my other LoRAs.
🔞This model's dataset contains NSFW materials.
The Pony V0.1 version was trained on only 28 images with rushed tagging. The Pony V1 version is the first one I consider "proper".
None of the pictures in the datasets had watermarks/signatures/artist names (although a few have text), if you get them - they're NOT from this LoRA's dataset. There may be bias towards certain elements.
Description
IMAGES
The number of images has been increased to 54.
To be clear, while I HAVE drawn worms with pecs/breasts, arms and legs, the images in this dataset (and for the previous version) don't have any such details.
TAGS
Removed tags "limbless", "armless", "legless", a bunch of needlessly complex tags, and age rating tags that I probably shouldn't have used for tagging to begin with.
This dataset doesn't have the tag "bald", as worms are bald by default. Any picture with hair is tagged "hair" and descriptive tags for it.
Added gender tags to most of the dataset, I prioritised genitals but mostly followed gender stereotypes ("female" worms are more likely to have hair and eyelashes, eyelashes are also tagged).
Hands are now tagged with "worm hands" and "worm hand".
Most worms have (visible) eyebrows in this dataset, so it doesn't use simply "eyebrows" but rather descriptions of them. Worms without (visible) eyebrows are tagged "eyebrowless" and "no eyebrows".
I added tags "g2worm" and "wmdworm" to a single digit number of pictures each. Due to low amount of / mixed style data they may not work well or at all. "g2worm" is used for any pictures with a worm drawn similarly to the G2 game style (introduced in Worms 2), "wmdworm" - W.M.D style, and while the overall style could be recognised, the technique didn't always match.
This one is weird... I added tag "inconsistent style". There are some pictures where different versions of worms interact, there may be differences not only in the shape of eyes, eyebrows and hands but even the way they're drawn. I don't know if it does anything.
Many other tag edits
There are some pictures in this dataset (they were in the older one too) that are too complex to be easily tagged Booru style, but writing essays just to describe them would be way too much hassle and I can't trust AI with Worms.
Trained against Pony Diffusion V6 XL via Civitai trainer on July 9. I also increased the number of Epochs this time.
Details
Files
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.