Lora trained on screencaps and promotional art of Tootie from Fairly OddParents.

I used the same tagging method as the Tootie's Mom lora, but the results were far less optimal this time.

After some testing, it would seem the trigger word failed to learn the subject, so it needs more tags to get results consistent. At minimun, include "black hair" as a tag (it can at times give other hair colors depsite none being present in the dataset).

Use "toon (style)" to better match the show's artstyle, but keep in mind there there will be more errors this way. The lora has a really hard time learning her eyes and nose and will duplicate them. If "twitnails" is not specified it might give her a ponytail instead, but the inclusion of the tag can also result in tri or quad tails.

Lower the strength of the lora to reduce the chance of errors.

This was my second attempt at a Tootie lora. The first one had stronger artstyle but the same eyes/twintails error and the results a bit less controllable/predictable.

Quite frustrated with how tricky she is to train, not gonna lie.

I'll probably do a third one at some point to try and iron out the unexpected mistakes in this dataset. Maybe include images of her mom or train the two together since they share design elements?