PLEASE DO LEAVE A COMMENT/REVIEW/LIKE. It helps in me earning Buzz so that I can train more. if you have any specific requests for any LoRa, DM me on here

V0.2 - The Raw Trained Version

V0.3 - Retrained from Scratch with different dataset based on feedback from V0.2

V1.0 - Size Reduced version:

I did multiple tests with my face similarity checker tool and found that the Loss in quality is somewhere around 19.69% over the Originally sized LoRa(V0.3) for individual test pictures. For most part, the outputs are pretty similar to the naked eye with the advantage of reduced Size of the LoRa.

Trigger word is : ssmaheshbabuukohwx

suggested settings :

"model": "flux1-dev",

"steps": 30,

"cfgscale": 1.0,

"width": 1536,

"height": 1024,

"sampler": "uni_pc",

"loraweights": "1"

"cliplmodel": "clip_l.safetensors",

"clipgmodel": "clip_g.safetensors",

"txxlmodel": "t5xxl_fp16.safetensors",

"negativeprompt": "",

"vae" : "ae"

Please note that this is an experiment and This model is purely trained for educational and research purposes. This is only meant for SFW and ethical image generation.

This LoRa was trained with 62 images of various angles of face, poses for 10 repetitions for 10 epochs