CuteCore

CuteCoreXL is a model that has been based on Animagine for v1, autismmix_confetti for v2 and 4th tail 0.3 for v3. The model is merged with over 30 personally trained LoRAs in an attempt to achieve the cutest styling possible. Many of these LoRAs have been made public, just not here because it's a lot of work to upload.

I have left this labeled as a base model because the original CuteCore for 1.x is, in fact, trained. Please see the original description below for information on that which is unrelated to the new XL model. CuteCoreXL is not suitable for training or merging, but you're welcome to try!

Recommended negative:

nsfw, simple background, censored, blurry

That's it! No need for any fancy negative with this model. Please do not use style tags such as masterpiece, best quality or quality tags such as quality_9 with this model. It is also unnecessary to negative source's as is often required with anything pony based.

Original Description:

This model has been fine-tuned on over 75k images of cute characters from a base of NAI on NVIDIA A100.

This model should not be used for ???? purposes but if it is I will not be held responsible.

This model should not be used for further training as it contains data that no other model does and would break cross-compatibility.

This model has been tested for compatibility with character LoRAs trained from numerous other models successfully while retaining style.

This model has been tested for compatibility with concept LoRAs trained from numerous other models successfully.

This model has been tested for compatibility with style LoRAs trained from numerous other models with varying degrees of success, please use a lower weight.

Due to the training base this model works best with LoRAs created from NAI.

All previews are generated with only this base model - No LoRAs, no Textual Inversions, no Aesthetic Embeddings, no ControlNet. The only exception to this is the last preview which used a LoRA to show compatibility.

All previews are generated without negative prompts. All previews are generated at 512x704 with 2x hires fix. using the latest WebUI and 4x-UltraSharp upscaler.

Great care has been taken to ensure that previews are within the bounds of the rules and that there is ample description provided of what this is and how it was made and used. This model should not be eligible for deletion as all rules have been followed. If any previews are found to be ???? in any way - please seek help, you need it.

The traits of this model include specifically cute and adorable looking with a focus on correct body composition without need for negative prompts. Heterochromia was specifically trained as well but is still random colors only. Negative prompts may still be used if needed. Hires fix. is recommended. The better your prompt the better the output.

This model may or may not be capable of creating older or more mature looking characters. It has been attempted for a short period of time with low success rate.