If you enjoy my work, please click the thumbs up. Thank you! /Bow
>>> UPLOADING/SHARING MY MODELS OUTSIDE CIVITAI IS STRICTLY PROHIBITED* <<<
AGDV2_nokrea is very much an experiment. Back in the day, when using SD 1.5, I occasionally created a beautiful image that I could not replicate. I started placing those images in a special folder in hopes that, at a later date, AI would advance to the point where I could create a lora with them using a single image.
This is now possible, although the technology is still in beta.
Admittedly, I have a fascination with the 1950s in America. I love the aesthetic, clothing, etc. AGDV2_nokrea was created when I was going through my 1950 retro diner phase using the versatile sd 1.5.
The Flux version was created from the following image. 
Description
FAQ
Comments (4)
what is the method ? your samples are very nice
First and foremost, my prompts tend to be rubbish, leading to many happy accidents. Being serious, it's a combination of imagination, trial and error, and borrowing what works from images others have posted.
When an image has no generation data, I run it through a program called Dataset Processor Desktop; this works best in SD 1.5, sdxl, and Pony. That being said, sd 1.5 prompts also work with flux.
I also use a Flux prompter program to generate flux prompts from images that interest me.
I hope this answers your question.
I hope you have a fantastic day/night. /Bow
@KelevraQuakenstein very great thank you ! and how did you train your lora with only one picture ? random cropping, flipping etc ? or an online service ?
@magnusfolk314971 I use a comfyui workflow designed by a Patreon creator and occasionally an online service.