Big Macromastia Breast LoRa for Hun | Wan Video
Trained at 211 on a dataset of 889 clips 1sec to 3sec (several source videos), with resolutions from 256 to 512, 5e-05 Learning Rate, Num_Repeat = 2
The workflow im using is:
Hunyuan 🌻 AllInOne ✔All Versions 𝟭.𝟱🟩🟧🟪 - Ultra1.5 from @LatentDream
IMG to VIDEO simple workflow WAN2.1 | GGUF | LoRA | UPSCALE | TeaCache @UmeAiRT
TXT to VIDEO simple workflow WAN2.1 | GGUF | LoRA | UPSCALE | TeaCache @UmeAiRT
* you can find a slightly modded wan workflow inside the sample video's metadata
Example prompt:
{an asian woman|a black woman|a woman with fair skin} with M4CROM4STI4 wearing a {military suit|business woman suit|flight attendant attire}. Her skin is fair, and she has {long wavy brown|bobcut|pigtails|ponytail|short} {crimson|brown|white|blue|blonde} hair. Her breasts are gigantic. She wears minimal makeup, accentuating her natural beauty. The lighting is soft, highlighting her curves and creating a sensual ambiance. Her expression is sultry, with slightly parted lips and a confident stance. She {sways her hips|dances|squats}. The background is dark, drawing focus to her.A beautiful woman with large m4crom4sti4 breasts and dark bobcut hair, stands nude in a softly lit room, her fair skin glowing. She has big blue eyes gaze directly at the camera. Her breasts are full and natural, with pink nipples. She has a slender waist and a flat stomach. Her pose is relaxed, with her arms hanging loosely by her sides. The background features a white wall and a wooden dresser with a green plant on top. The camera angle is eye-level, capturing her from the waist up.a woman with m4crom4sti4 with large breasts, wearing no clothes, standing in a bathroom. She is facing away from the camera and is seen tying her hair up with a hair tie. The bathroom has white walls and a white door. The woman's body shape is curvy, and her breasts are prominent. The movement of her breasts is minimal as she ties her hair. The woman's position is upright, and she is facing the door.Have fun! 😎
***** This LoRA is shared without warranties and with the condition that it is used in a lawful and responsible way. I do not support or take responsibility for illegal, harmful, or harassing uses. By downloading or using it, you accept that you are solely responsible for how it is used. *****
Description
FAQ
Comments (12)
I really hope to get your training data, so that as a newbie, I can also train my own Hunter LoRa
You meant hunyuan loras? XD
Im using this for training on a 4090
https://civitai.com/models/1085714?modelVersionId=1219185
Also you will need a video captioner for the videos you wanna train with.. there's plenty I'm using https://github.com/cseti007/LLaVA-Video-7B-Qwen2-captioner/pulls
hang n swang
Any chance we can get a WAN T2V version?
I've heard that works with vace wan models as t2v
@K3NK from what i can see from trying and what i can see online, VACE is only compatible with T2V loras.
just trying it on comfy with vace gave a "lora key not loaded" error meaning it's incompatible.
@Griphen116 please see my posts attached to this lora, I use it on t2v with great success :)
I can confirm it works with vace, though there are some lora loading errors in log.
Also seeing lora loading errors with FusionX base model T2V but is still producing great results so I guess they don't matter.
Does a loading error negate or diminish the LoRA effect? When I see any loading error I usually remove the LoRA assuming it is incompatible.
On another note, t2v is great for generating Images, the more t2v LoRA the better.
Using the WAN version now. Still the best for large natural breasts. I find the best ratio of effect to video quality preservation to be 0.8 strength.
I also want to note that you misspelled the single word trigger on this page. It is "m4crom4sti4", as can be seen is the sample prompts.
However, sadly, even at 0.4 strength, I'm getting lower motion adherence on position loras if I use it.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.