I am happy to present the 1st iteration of my HMFemme LoRA!
A LoRA aimed at achieving realistic looking female in candid every day poses with realistic lighting and shadows.
Join the HearmemanAI Discord for support, workflows, exclusive tips and tricks and more!
This LoRA was trained on a large dataset using my Diffusion Pipe RunPod template using an H200.
Best practices for HMFemme:
- Performs the best with character LoRAs.
- Use descriptive and elaborate prompting to achieve the best results, more on the below.
- Perform very well with my Qwen Breasts LoRA as it was also designed with a realistic candid approach.
Prompting - Please READ
I used very elaborate captions to get results as flexible as possible.
Prompts should be descriptive and always start with "HMFemme, an amateur photo taken from a smartphone camera"
Please refer to the example images to see the prompting style.
Workflow:
I recommend using this workflow.
It's also implemented in my Qwen RunPod template.
For members of my Hideout community, I've developed an AI Agent for creating HMFemme prompts, the agent is NSFW compatible and accessible in "The Hideout" category on my Discord server.
I hope you enjoy using this LoRA
Description
FAQ
Comments (15)
This looks so good holy shit
Thanks!
Is a 5090 necessary for this workflow because i have a Asus 4090 and it eats 99% of my VRAM
Try nunchaku version for qwen
Great Lora! Thank you for your efforts.
Would this also work with 2509?
Haven't tested, feel free to test and report back
Sadly doesn't work good with Lightning 4steps lora.
Almost nothing works good with that LORA
@_Dream_Making_ not true.
Try playing with strength settings for these loras. Only lighting is set to 1. The other loras are rarely set to 1🤞🏻
lightning 4steps lora seems need a very short and 1 task edit ...other than this will be a mess
The current world population is approximately 8.23 billion people as of 2025. It is projected to continue growing, reaching around 9 billion by 2037.
now introduce ai people........lmao hahahahahaha
good job, we need more people bored and focused on themselves............lmao ..oh the trends.. as i work on a illus realistic base model because of this insane trend to want to make more people... lmao
I'm interested in 1boy alternative. There are already too many 1girl and almost none 1boy loras :-).
great lora but can u pls mention what weight and samplers to use with how much step and i found a workflow in ur discord and i cant find how i can create multiple images casue it always generates single image and i'm using fp8e4m3 version cause bf 16 is very large and i'm getting incorrect & blurry results so is it not gonna work with fp8e4??