SD3.5L Naytlayt ???? Style LoRa

What is this?

The base SD3.5 Large model lacks training data for ??????, which is a hinderance even when doing SFW prompts.

To solve this , the 'Naytlayt' LoRa serves as a stopgap measure to "fill in the blanks" in the SD3.5 for female ?????? and ???????.

Best usage for this LoRa:

Pick an image online (something like a magazine page will work well) run it through JoyCaption Alpha One , and use the prompt: https://huggingface.co/spaces/fancyfeast/joy-caption-alpha-one

Above examples is from using the SD3.5 Large base model. For better consistency , use a fine tuned SD3.5L model if available.

This LoRa will give different results when used on the SD3.5 Medium model or the SD3.5 Turbo model.

//---//

Training method:

This LoRa was trained on SD3.5 Large using fp8 precision.

Training images are 45 images of ????? female anime characters on a single color background.

All training images only feature a single female character.

As a result, groups may be "clumped together" when using this LoRa.

As you can see in this chart, even with different characters and perpectives , the loss rate is low:

Only a single training image in this set uses a white background. The rest are of different colors, set in a photo editor: blue , red , green , yellow and black.

The purpose of this step is to prevent the SD3.5 Large model from creating "pale" looking images with a lot of white in them.

//---//

Feel free to use this LoRa however you like. You may merge this LoRa into your own LoRa project and/or model without crediting me.

//---//

Prompts for training images are created using Joycaptions: https://huggingface.co/spaces/fancyfeast/joy-caption-alpha-one With the keyword 'naytlayt' added close to the start of the prompt.

All training prompts were kept between 550-800 characters in length. The reason for this limitation is to keep the text prompt within a single 256 batch encoding for the T5 encoder.

To quote Stability AI:

"While this model (SD3.5) can handle long prompts, you may observe artifacts on the edge of generations when T5 tokens go over 256. Pay attention to the token limits when using this model in your workflow, and shortern prompts if artifacts becomes too obvious."

Source: https://huggingface.co/ckpt/stable-diffusion-3.5-medium