(https://www.fiverr.com/partai/develop-your-custom-lora-or-loha-model-for-stable-diffusion?utm_campaign=gigs_show&utm_medium=shared&utm_source=copy_link&utm_term=q74apzy) (https://ko-fi.com/part_loras)

This was a LoRA made for a commission! My client was kind enough to let me share on CivitAI; thank you!
Edit: I saw that he/she posted it already -> https://civitai.red/models/2841647/pole---

No trigger words! Just use the LoRA, suggested weight is 1.0-1.5.

Start your prompt with: "pole , collar, ...". You can also include "anime screencap", "2d", "flat color", "illustration"; highly recommended but not required! Read below to understand why.

The LoRA was trained on a SYNTHETIC dataset created with Krea 2. The original dataset was frames from a real world video that I turned into anime-ish images using Krea 2 + ControlNet. My client really wanted the "thigh strap around ankles" (like the first sample image) to show up in squatting images so the dataset was optimized specifically for that, anything else is just an extra bonus.

ComfyUI Metadata is included in the sample images! Click the little icon in the bottom right to see it.

It was trained on 50 1024x1024 images scraped from the internet, extracted from a video and style transferred to Anime using Krea 2.