NOTICE: Some version may have trigger words while others may not.
I improved my knowledge on LoRA training, train alot, and here we are. Vpred version didn't came out well though so no releasing that,sorry (Vpred is dead anyway).
........new Noob EPS news soon wink wink
Anyway this style is one that I've wanted to create a lora of for a long time, the artist is incrediby skilled at what he does. This also means it has been a harder training period compared to my 1st lora.
All the showcase gens are also RAW AND NON UPSCALED , I want to be as truthful of the lora quality and result as much as possible.
Suggested Tags:
YHoundbell, holding sword, black gloves, parted lips, long hair, black hair, very long hair, black pants, black jacket, blue eyes
Disclaimer
I respect the artists who honed their crafts for years to achieve these styles among all else, if any artists whom LoRA i've made from shown displeasure/objection to their arts being used as AI training material or if they request a takedown notice on Danbooru , I'll delete the LoRA with no further notice.
Description
Trained on the CabalResearch/NoobAI-RectifiedFlow-Experimental · Hugging Face
V0.2 model by Bluvoll and Anzhc, the model is based on Rectified Flow and EQVAE, which is superior to eps/vpred and sdxl vae, and through my training experience is also confirmed to converge faster and better/more accurately with LoRA training. I hope more people will begin to notice CabalResearch efforts in improving SDXL, which still have much to give.
Oh right I'm training Rectified Flow model by using 67372a/LoRA_Easy_Training_Scripts at refresh
You can also train on sd-script CLI with Bluvoll fork bluvoll/sd-scripts
Seek me out through discord if you're interested in how to train Rectified Flow loras (talannnn).