Footgagging LoRA for Wan 2.1 T2V 14b

The previous model has been fine-tuned. Vertical videos have become a little better. Adjust the Lora weight to your liking.

Model Type: LoRA
Base Model: Wan2.1_T2V_14B_FP16
Sample Outputs: Generated with Wan2.1_T2V_14B_FusionX
Recommended Use: Video generation (T2V)

🔧 Recommended Settings:

• LoRA Strength: Use 0.55 or above. Adjust per pose for best results.

• CFG: 1.0

• Shift:

  • For realistic/semi-realistic: 2

  • For anime/cartoon styles: 8–9

• Resolution: Best results at 864x480 (16:9), but other resolutions may also work well.

📌 Description:

Footgagging, this LoRA is trained to generate video scenes where a female character inserts her foot into another character’s mouth. It focuses on pose clarity, foot insertion realism, and emotional contrast (dominant vs. submissive expressions).

Works well with both realistic and stylized characters, and supports a wide range of scenarios—from fantasy to modern.

📷 Prompting Tips:

• Always start prompts with: Footgagging,

• For foot insertion, use phrases like: inserts her foot, is inserting her foot, inserted a foot deep, etc.

• Example prompt style:

  Footgagging, a smug elf queen seated on a crystal throne, is inserting her bare foot into the open mouth of a kneeling guard.

Using "insert/inserts/inserting" expressions ensures consistent foot-in-mouth pose results across keyframes.

It may be a bit weak in producing different poses since it is usually trained with similar dataset. When I tried to train different poses at the same time, it failed. But I am working on fine-tuning it now. I hope it works.