A LoRA trained off of the works of Etlabsotwe (https://twitter.com/etlabsotwe?lang=en).
Two versions are available, one trained on animefull-final-pruned and one trained on AOM3A1. The AOM3A1 model is much better at capturing the style, but the animefull-final-pruned model can produce good expressions while maintaining the style of whichever mix you choose to use it on.
UPDATE:
Added version trained on Pony Diffusion V6 XL.
UPDATE 2: Added version for Noob. Trained off of Noob EPS 1.0, previews generated with IllustriousXLPersonalMerge.