Red Spider Poncho, Bandit and Mage outfit - 3 in 1 lora. All images are generated with A1111 on PC. Trained with Kohya with 40 pics and a realistic checkpoint and tested with my Illustrious model available on TA/HUB. Images for training made with some anime lora available on CivitAi and converted in realistic version with img2img for better realistic results. Thanks to the original creators for the anime version of the loras.

This model does not use any real-life or celebrity content. Please do not use this model for any real-life, celebrity, or pornographic content. (C) for QoS mark of the original manufacturer, not for commercial use.

Illustrious Weight 0.5 - 0.6 CFG 5 on realistic checkpoints. Check info below.

IMPORTANT: All my models work only if the trigger words and the recommended negative and positive prompts and embeddings are used. If you have a different result from the images, it is because you have not used the recommended settings. Too many loras and/or embeddings can ruin the model's result. To have images like the ones I made, use ONLY the recommended settings. If you use the indicated settings (below), you should be able to create images like the ones shown in the photos/videos. The delicate images with an ethereal gradient are intentional and included in the lora. With anime models, it could give completely different results. In the images, I used my semi-realistic model available on TA..

Example of positive prompt with trigger words:

<lora:3outfit77-000013:0.7> a beautiful girl wear a (red spidernet poncho), black fishnet thighhighs, oversized poncho, hip vents, scoop neck, long sleeves, off shoulder, poncho spider net print, black lipstick, soft makeup, big breasts, sagging breasts, sunny day, skinny body, big thighs, large hips, skin pores, mole, eyelashes, happy, look at viewer, pose, hairstyles, haircolors, background

<lora:3outfit77-000013:0.7> a beautiful girl wear a (bandit outfit), red gloves, striped crop top, striped skirt, striped thighhighs, red necktie, hat, black cape, navel, midriff, big breasts, sagging breasts, sunny day, skinny body, big thighs, large hips, skin pores, mole, eyelashes, happy, look at viewer, pose, hairstyles, haircolors, background

<lora:3outfit77-000013:0.7> a beautiful girl wear a (witch costume), black bikini, wand, witch hat, detached sleeves, neck bow, black thighhighs, bikini armor, short cape, bikini warrior, navel, holding staff , big breasts, sagging breasts, sunny day, skinny body, big thighs, large hips, skin pores, mole, eyelashes, happy, look at viewer, pose, hairstyles, haircolors, background

For back views add in positive prompt "from behind, looking back, ass" and delete "cleavage"

In negative prompt add "(nipples)" if you don't want to see this details.

Settings For Illustrious version:

IMPORTANT, PLEASE READ : If you get different results, the prompt was probably not followed and an unsuitable checkpoint was used. All my LoRAs have been trained on realistic models. The videos I post represent what you can achieve; even though the video is animated, it reflects the quality and what the LoRA can do. The first frame of the video is the image I created using the LoRA and then animated, using the parameters I recommended. To achieve a realistic result, a realistic ILLustrious non-merged checkpoint should be used (or Semirealistic based on realistic checkpoint) . If you use merged models with anime models, the result may be very different from mine. Do not use corrective loras in addition to the main lora because they often only ruin the result. If possible, do not exceed 150 tokens in prompts. Too many words in prompts do not improve the result.

Do not use embedding or corrective loras, they can compromise the result. Too many loras together are not good. The best result is obtained with one lora or two and a prompt with less than 100 tokens,

The parameters I use are the following :

Steps: 35+, Sampler: Euler a, Schedule type: Automatic, CFG scale: 4.5, Size: 512x768 or 768x1024, Denoising strength: 0.35, ADetailer model: face_, ADetailer confidence: 0.4, ADetailer dilate erode: 4, ADetailer mask blur: 10, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 64, ADetailer use inpaint width height: True, ADetailer inpaint width: 768, ADetailer inpaint height: 768, Hires upscale: 1.5, Hires upscaler: 1x-ITF-SkinDiffDetail-Lite-v1,

For Adetailer face, you can choose the better: face_yolov10x_

Reposting on other sites is prohibited.

It is strictly prohibited to distribute outside TA-HUB this model or hosting services in third-party image generation services, or sell my images on other site. You can use this model for personal use on HUB or A1111 on your pc. Not for sell or repost on pixai or other sites.

If you appreciate the lora, please consider leaving us a review and sharing the images generated using this model. Your feedback is valuable to both us and the AI community. Thanks.

Halloween/Carnival collection 1 - 3 in 1 lora - v1