Join the NSFW AI community on Discord where over 6,000 NSFW AI developers, creators and enthusiasts gather to share tips and tricks, talk about the latest models, and collaborate on research to push the cutting edge of NSFW AI!
https://discord.gg/mjnStFuCYh
**Introducing Breast Massage Therapy by BetterPorn**
---
### What is it?
**Breast Massage Therapy** is a pose/concept LoRA designed to generate artistic and serene images of topless women receiving gentle breast massages. This model focuses on capturing the relaxation, comfort, and tranquility associated with therapeutic massage practices.
---
### How to use it
**Important:** This model is versatile and works well in various settings and orientations.
- **Trigger Word:** To activate this LoRA, include the trigger word `"breast massage"` in your prompt.
- **Strength Setting:** Start with a LoRA strength of **1**. Adjust as needed, especially if you're blending with other LoRAs.
**Getting Started:**
Begin by outlining a simple "who, where, what" in your prompt. For example:
- `"A woman, serene spa setting, receiving a breast massage "`
**Examples:**
- `"A woman with long, wavy brown hair lies on a massage table bathed in soft candlelight, her skin glowing as she enjoys a breast massage"`
- `"In a tranquil garden setting with blooming flowers, a woman with a peaceful expression is massaging her own breasts, sunlight filtering through the trees"`
Feel free to enhance your prompts with details about the environment, the woman's features, and the overall mood to create a more vivid and personalized image.
---
### Mixing with other LoRAs
**Breast Massage Therapy** is highly adaptable and works well when combined with character or style LoRAs to achieve a specific look or artistic style.
- **Character LoRAs:** Integrate with female character LoRAs to depict specific appearances or personalities.
- **Style LoRAs:** Mix with style LoRAs to apply different artistic techniques or moods to your images.
**Tips:**
- Adjust the strength of each LoRA accordingly to balance their influences and achieve the desired outcome.
---
Enjoy creating calming and beautiful imagery with **Breast Massage Therapy**! Let your creativity flow as you explore various settings and styles with this specialized model.
Like this model?
Check out our discord https://discord.gg/mjnStFuCYh
We're a community of AI developers, creators and enthusiasts collaborating, sharing what we learn and generate, and pushing the boundaries of what's possible.
Description
FAQ
Comments (13)
only 77mb? damn, need a tutorial haha. Looks awesome
When I discovered that Hunyuan would generate pretty much any explicit NSFW content with the right prompt, I determined that LoRAs only need to very slightly push the model towards consistency of that concept. This was trained with 8 dimensions, most are using 16, 32, or more but I don't think that is necessary in most cases.
I do have a guide in fact for exploring the boundaries of what the base Hunyuan model will generate with prompts alone: https://civitai.com/articles/10484/enhancing-content-generation-with-hunyuanvideo-using-comprehensive-prompts-and-negative-prompts
This is impressive.
I know you have your prompting thesis article but it might be wise to make a lora creation guide as well. 77mb is insanely optimized. People need to learn this way so they stop making 800mb loras
It's just a rank 8 Lora. Most Loras are rank 32 on this site.
Lower rank Loras have less influence over the original model, so if the original model can already get pretty close to your desired prompt, you don't have to create such a beefy Lora.
For characters, clothing, and complex actions it's generally recommended to use rank 32 Lora to capture more detail. This is usually what results in a 300MB file.
Does not work for me somehow. The only lora that doesn't work. I did redownload. Restart. Using Hunyuan FastVideo fp8
lora key not loaded: lora_unet_double_blocks_0_img_mlp_fc1.alpha lora key not loaded: lora_unet_double_blocks_0_img_mlp_fc1.lora_down.weight
[SNIPP]
lora key not loaded: lora_unet_single_blocks_0_modulation_linear.lora_up.weight lora key not loaded: lora_unet_single_blocks_10_modulation_linear.alpha
[SNIPP]
Did you use https://github.com/kohya-ss/musubi-tuner for lora training? Maybe it's related to it
@lost_moon Yes I did
@BetterPorn You've probably seen it already, but I'll post it here too just in case someone else will have the issue. The solution for this issue is the following, with musubi. This lora needs to be converted to work with comfyui properly. After doing so I get no error anymore and the lora is working:
python convert_lora.py --input /pathto/lora.safetensors --output /pathto/lora_output/converted_lora.safetensors --target other
@lost_moon How did you convert it? I have my LoRA from musubi and yes it got error when I used it on Comfy,..
@YesPleaseProceed You got to open a cmd command window in your musubi folder containing the convert_lora.py and enter the command I posted in the other comment. Replace the pathes with your paths, as well as the file names with those of your lora. Hit enter. That's it.
@lost_moon Thanks mate!
@lost_moon Hey Mate, yes it works! But my LoRA quality is poor, apparently. I prefer to train with diffusion-pipe rather than musubi tuner. IDK, but with low resolution like [244] or [256] and around 40 epochs, the LoRA results is amazing.
Woa, just discovered hunyuan. Cant wait to test your lora. :D