Woot Art Hard Shading Oekaki Style TI Embeddings

This is two different textual inversion embeddings, so be sure to try out both (as well as both together, they are compatible). They bring out the best in the related LoRA's style, dramatically improving style fidelity.

Both embeddings use the same training data as the LoRA. The numbers are from a longer series of experiments, but these two are the only ones worth sharing.

woot-style-06 is trained on Illustrious v1.0 without the LoRA. As such it kinda works without the LoRA present, but when paired with the LoRA they build off each other. If used with the LoRA you should include the trigger word ws drawing as normal as the embedding doesn't include it. Generally speaking it makes gens that are more cartoony and makes the ladies bigger and rounder.

woot-style-09 is trained on Illustrious v1.0 with the LoRA. Since it was trained with the LoRA, you MUST use it with the LoRA to get good results. Note that the embedding effectively contains the LoRA trigger word, so you can omit ws drawing if desired, but it's probably a good idea to use it in your prompt anyway if you have the tokens. Generally speaking it accentuates the woot style, though it also has some bias towards bigger ladies as well.

Also note that if the style is too strong you can reduce the weight in the prompt. If you don't want bigger ladies this can be useful, though prompting is usually a better solution. It also generally has a poor impact on hands due to the input data, so this can help with that issue. That said, a weight of 1.0 on the embeddings is usually fine.

Note that these are POSITIVE embeddings, intended to be used in your positive prompt.