FLUX V1.0
Mixed the "Rebirth" pics to the SDXL1.0 dataset
Base model: FLUX.1 dev
Added the outfits from "FFVII Advent Children" and "FFVII Rebirth"
Weight: 0.6-1
Using it with Adetailer for better face generation
Outfit prompt (FLUX only):
Default:
black sports bra,white crop top, leather suspender skirt, elbow gloves,,arm guards,arm pad
Advent Children:
red arm ribbon, black gloves, leather crop top , midriff peek, navel,zipper,sleeveless,black knee height shorts, leather skirt
Cheongsam:
grey cheongsam, short hair,double bun hairstyle, cleavage cutout, bracelet, jewelry, leopard print, corsage, high heels
Kimono:
black thighhighs, short kimono, japanese clothes, black kimono, black hair, sash, obi, boots, long hair,long sleeves, black boots, wide sleeves
Purple dress:
purple short dress with black outline, jewelry, crescent earrings, cleavage, sleeveless dress, bare shoulders, black neck ribbon, purple heels
Serafuku:
white school uniform,red neckerchief, bracelet, white socks, pleated skirt, standing, long sleeves
Shining spirit:
white frilled bikini with dark blue stripe, hair flower, hair ornament, navel, cleavage, wristband, frills skirt with dark blue stripe, short hair, messy hair, costadesol-hairstyle
Majestic Glamour:
denim shorts,black striped tied shirt, sleeve rolled up, navel, belt, midriff, underwear, highleg panties, cleavage, crop top,, black shorts, short hair, messy hair, costadesol-hairstyle
Cowboy:
cowboy hat, cowboy boots, black gloves, brown leather skirt, midriff, belt, vest, cowboy western, navel, orange skirt, fringe trim, bracelet, sleeveless,
SDXL V1.0
Base model: SDXL 1.0
Weight: 0.5-0.7
Using it with Adetailer for better face generation
The ancestral sampler seems perform better
Any comments or suggestions are welcome.
Description
FAQ
Comments (16)
I just copy your prompts and the photo is broken, all the setting like yours.
Might be the VAE or the sampler problem
And I'm testing it on the SDXL_fixedvae_fp16 checkpoint, you can also give it a go.
Thanks, is the VAE problems
have a flavor inside
Hey your Tifa AC model is the best IMO, would you consider doing a Yuffie AC model too? If she doesn't have enough scenes you could use some from dirge:)
Sorry for the late reply, been very busy these days. The problem is that the Yuffie model looks different in AC and Dirge. Even within the same movie, SE tends to only use the high-poly model in the close-up scenes and the mid and low-poly models in other mid and long shots. The same thing happens with the AC Tifa, but she has more close-up scenes to compensate for the differences. For now, I can only capture around 30 acceptable screenshots of AC Yuffie, but that's not that enough to make an acceptable Lora model
absolutely agree this is best Tifa lora
Thank you. This is the first tifa Lora I found for SDXL. And it works fine.
I'm still trying. SDXL is not easy to control
I want to know what your vae is
hi i use easy diffusion do you know how i can get DPM++ 3M SDE?
Can anyone get something solid from this lora? I've tried downloaded SDXL suggested as well as the vae. Tried low lora weight, higher clip, low cfg, change sampling, etc. I either get too much noise artifacts or just noise. Can't get anything close to the FF7Rebirth model. Would appreciate suggestions on getting this working. Using comfyUI btw.
I just uploaded a batch of pics using ComfyUI. You can try importing those into your setup.
@Monobot Thanks for the node info. I tried using it and had to modify some stuff (default language settings are different I think) but I'm getting pretty close. The face is still kinda... gremlin-y lol. I'll keep at it.
Can you do Jessie and Aerith too?
The best model , thank you very much!