The style of Spidu.Trigger word is <code>spidu</code>. Consider adding <code>censored</code> to the negatives if you encounter censored genitals. The model should do well at <code>excessive cum</code> and <code>piss</code>/<code>pee</code>. Also, the female characters from My Hero Academia were prevalent (and tagged) in the dataset, so they might also work without specialized LoRAs.Trained on ~250 select images. This time I decided to try using @THEANTLERS' recommended approach (used fewer images of higher stylistic consistency, used WDv3 autotagging, added <code>style_9</code> to training tags and used the params from their config). I think I'm happy with how it turned out, I might even revamp some of my other LoRAs this way.I'm also going to try early access with this one, if I can figure out how to get it working. If I manage to get back the buzz it cost me to train this, it'll make it significantly easier to train more LoRAs in the future.If you like the style, please support the <a target="_blank" rel="ugc" href="https://x.com/spidulewds">original author</a>EDIT: Well, I couldn't figure out early access. Maybe it isn't available to everyone just yet? Not a big deal, I guess, but this means I gotta save up some more buzz before moving on to other styles.18.03.2025 - Updated to IllustriousXL. Consider puttting <code>censored, mosaic censoring</code> in the negatives.

Spidu style (ILLUST & PDXL)

Update for Illustrious. Same dataset. Has a tendency towards censoring private parts, consider putting <code>censored, mosaic censoring</code> in the negatives

Files with SHA256: 6f83bb1f38626409e4c0b5936c9d41c3fb69774ac7cc82948fdafa14f0cc57c3