PhotoVintageV1.5 - 512x512
SD 1.5 trained model with assorted photographs from my work as a professional photographer. The photographs were processed with our own color filters (LUTS) to apply one of my favorite styles (vintage style).
This model is focused on photorealistic images. It can spawn all sorts of objects and people with quite an interesting quality.
I recommend using:
- 30 - 40 steps - 7 - 13 CFG
- euler a, ddim... (Experiment)
I RECOMMEND USING THE FOLLOWING PROMPT:
vintage polaroid analog portrait photography of a (HERE YOU CAN PUT THE SUBJECT OR OBJECT YOU WANT), warm azure tones, heavy lensflare, color bleed, film grain
"vintage polaroid analog portrait photography" (They are the key words to achieve the photorealism of the images.)
"warm azure tones, heavy lensflare, color bleed, film grain" (They are the key words to achieve the vintage colors and finishes of the images).
Feel free to experiment with keywords, styles, and whatever your imagination can afford with this model.
As a final recommendation... You can use references of photographic cameras, photographic lenses, focal lengths, diaphragm openings and any photographic characteristic to make your creations more stylized.
HUGGIN MODEL FOR TRAINING: fotografomedellin/PhotoVintageV1.5 · Hugging Face
By: Jhoan Giraldo - Fotógrafo Medellín
Description
FAQ
Comments (12)
So fun to see people making models of their own work. It'll be fun to compare this to Analog Diffusion.
Do you feel like the model represents your work very accurately?
It manages to represent many graphic elements that I usually use in my photographs (colors, types of focus, styles...), so I was satisfied. In fact, he usually achieves things that I didn't even imagine for my photographs... So he feeds me back and serves as a source of inspiration.
@NecroBear you should check this out.
imagine merging them, considering they're both trained on OG datasets
why make it a checkpoint? Just make it an embedding.
Although an embedding usually weighs less and is more practical... I wanted the model to have more flexibility. So whoever uses it can not only take advantage of photorealism, but can also experiment with other styles in which photorealism will give interesting touches. In addition, the model hosted in Huggin, possibly serves as a basis for other people's training... And that's what it's all about, it's a collective advance. 😎
@fotografomedellin Thank you for this checkpoint, I've been having a lot of fun with it!
@fotografomedellin Awesome!
Do you know anything about the linked comment or model?
Comment: “This model produces the exact same pictures as the model by another user PhotoVintageV1.5” 😢
It is common for some weights to be similar to other models. Remember that almost all the trainings are done with the original SD 1.5 base... So if you train a "vintage" style model with your photographs (that handle the colors and textures of that photographic style), and I do the Same with my photographs... Most likely, there are similarities. Now... If someone says that they trained their model from scratch, without using any original SD base model, they are surely lying... Because that would be very expensive 😅
In the case of the model you mention... In fact, they uploaded it first. Mine is more recent... So I don't think it's a copy from them.
In the case of my model, it is based on SD 1.5 and photographs taken by me (I have been a professional photographer for more than 15 years. So I have a good number of photographs to train various styles). And the Vintage style is my favorite without a doubt. 😎
I still left a link to the model in Huggin in case someone wants to use it to modify it, improve it or train something new with my model as a base. Honestly, that doesn't bother me. In the end I like the idea that knowledge is collective.
@michelangelo By the way... I already gave you my opinion in the comment you just showed me. I hope whoever made that comment changes the rating of him because it's totally unfair.
Thanks for letting me know 😊
@klotz should remove or update the one star rating now that the question has been cleared up. Thanks for your sharing your contributions, clarification and perspective.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.