Projekt Melody (VShojo) | Original Appearance (4 Outfits) | LoCon

I have a request form

The Ai Hentai girl Projekt Melody in her original appearance from the Vtuber group VShojo . Works best around 0.8-1.0 weight . Training done on NAI.

Trigger : " ProjektMelodyOriginal , bandaid on nose, short hair, ponytail, hair ribbon, hairclip " always needed, " medium breasts " suggested for accuracy.

Suggestions/notes :

Leotard Outfit: " black choker, beltbra, underboob, harness, leotard, navel, white gloves, elbow gloves, white thighhighs "

Beltbra + Shorts Outfit: " black choker, strapless, beltbra, underboob, white gloves, elbow gloves, white shorts, micro shorts, white thighhighs "

Crop top Outfit: " halterneck, strapless, white shirt, print shirt, off-shoulder shirt, crop top, short sleeves, micro shorts, white shorts, white sneakers "

" halterneck, strapless " can be removed to mix it up.

" Elbow gloves, white gloves " can be added to get the look with them under the sleeves.

" White shirt " or " Black shirt " work for color changing.

Remove " White sneakers " if you want upper body shots.

Swimsuit Outfit: " black bikini, swimsuit, halterneck "

Sometimes gens funny bikini bottoms as the data wasn't the best.

She is very flexible clothing wise and NSFW able with almost every piece being removeable one by one so please go wild and have fun :)

Been a while due to some technical issues of my own doing but now I am back to make some more loras. Melody is one of my favs out there so I really wanted to make a great lora of her as the other ones of her left a lot to be desired. I split her original appearance from the newer Vtuber model to avoid bleeding issues so I will be making another lora for her newer looks down the line. I hope you all enjoy.

Feedback and reviews are always appreciated.

Nerdy training numbers:

Trained on D8Dreambooth trainer

Optimizer: 16bit AdamW

Training resolution: 768

Unet LR: 0.0001

Tnec LR: 0.00001

Unet weight decay: 0.024

Tenc weight decay: 0.030

35 Epochs - 6960 Steps

Trained on 175 images.