Identity-focused character LoRA for realistic portrait and full-body photography.
Trigger example:
aureliax, natural red-haired woman, freckles
If you want red Hair, specify it! It seems Zimage is overwriting the hair color with „generic“ brown.
Trained on a controlled 1024 studio dataset to ensure consistent facial geometry, natural body proportions, and stable anatomy across pose and camera changes.
Designed for:
frontal and 3/4 views
neutral studio lighting
portrait and full-body compositions
SFW and NSFW use
Recommended weight: 0.6–0.8
This is a core identity model. Styling, mood, and clothing can be layered separately.
Example generations are available in the gallery.
Feel free to share your results.
Description
FAQ
Comments (6)
Do you find it trains better if you don't have the character in different poses/environments/situations in the training data? I don't think I even noticed any behind views clicking through, and lots of near-duplicate head-on shots of her standing there.
That’s funny, in the same time I wrote a comment for V2 and what’s planned. Yeah many front views and portrait because I wanted to be sure Face is absolute stable for what I plan with her. From what I learned so far and as addition with Zimage: poses don’t matter at all, I’ll tried so much stuff already and the poses are variable as hell with the existing model. Only extrem angles are somewhat unstable if focused on specific body parts and hairstyles are to unstable atm.
Oh and forgot: since it’s only for character, environments would in theorie make her unstable in training.
Hey :) Thanks for trying or using my first Lora. Is there anything you’ll would like to be added? I already plan on V2: more stability with different Hairstyles, more stability with nudes and more backfires added to training data. I’ll Appreciate your feedback if you have some!
There are some errors in the anatomy, but I hope they will be corrected in new versions. The main thing is don't make her breasts bigger!
Nope, this was one reason I made her at first tbh xD Zimage has the tendency to do to big breasts. Thanks for the feedback! Could you specify where the errors happen? I had some failures with Zimage too even without Aurelia with „extreme“ put totally possible poses.