Dedicated to people who were born in the analog era
This is my first attempt at creating a LoRA model for SDXL.
I tried to avoid overfitting, but it's possible that training was stopped a bit early.
Feel free to test it and share your feedback.
Example prompt: 90s flash photo, lomography, analog photo, man and woman standing on the street, evening sky, half body shot, looking at viewer
I’m removing all my content from Civitai. Thanks to their ridiculous new moderation.
From now on, all my LoRAs will be on Hugging Face.
Description
added new concepts: light leak, lomography
FAQ
Comments (11)
This gave me just the look I was going for, thank you!
Glad you liked it!
Maybe it's just me but, lora (version 3) has 2 problems. 1) ocasional double head problem when generating in 512x768 resolution. 2) heavy bias towards portraits, dificalties/reluctance to unzoom.
(Using checkpont beautiful art v4)
You are right, there are problems in v3, but right now I don't have free time to fix it
Hi How did you generate the dataset to train this? i want to carry out such training for other models such as SD3 and SDXL to have a similar LoRa.
I did not generate the dataset. Most of the images were taken from Pinterest 🤔
hi @coyotle thanks for the reply,
so from what i understand, you captioned these images off pintrest and trained a LoRa?
can you share the dataset please??
@coyotle can you share the dataset please?
@arpitsahni04494 check your chat messages
Details
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.