This is a lora I meant to do for a while. It captures the essence of the 90s anime vibe. It also have a touch of vintage parting into it making it more of a retro 90s aesthetic then 100% anime. Like with faded colors and all.
Vintage, 1990s \(style\) add to your prompt for more effect, but you really don't need them.
Best keep under 1 weight, 0.5-0.75 being the sweet spot.
Best for SFW but can also do NSFW
Installation of the LORA:
Copy the safetensors file into the \stable-diffusion-webui\models\lora folder.
Description
Most recent version
FAQ
Comments (41)
cool style!!
More pony loras please!
Amazing job! How many pictures did you use to train this? Did you use an addon to train it?
35 pictures and (no addon?), just me messing around with some of the presets in kohya and tweaking some values.
@dalla1re Would you ever consider posting your dataset?
@Brouz I don't think it's a good idea since I do not have all the rights to the images so it's not my place to redistribute them... I will eventually make a V4 tho with more variety and better tagging
@dalla1re glad to hear that you'll actively improve this model, it's awesome!
Amazing model, Hope you'll update this in the future!
nice
GREAT
love the pony lora, looks really great, but please check your training set for jewelry and properly tag necklaces, choker, earrings and bracelets. The lora loves to slap these on often and negatives don't work, since it thinks these belong to the body :/
Will do! It was honestly put together very fast and didn't think it would do so well here. I might retrain it with better tags one day
@dalla1re The same thing happens to me, it's very difficult to get the medieval background and the sleeves on the arms, I haven't been able to do it for hours.
absolutely love this lora! Thanks for providing this!
Just two things: any good tipps to improve the hands? I mostly use it for anime stuff, and I haven't really figured out a way to improve hands (apart from the obvious negative prompts).
Also, the backgrounds tend to have a tendence towards "castle walls" / dungeon like setting? Any tipps?
=)
Thank you! and for the hands I don't have any tips unfortunately. For the background it's due to the dataset that has a lot of those in it. If you want something else, prompting it usually works. (like forest background, gradient background, etc.)
Really great work, I'm very much enjoying working with this LoRA. Forgive me if this is a basic question, but what does all the "score_9_up, score_8_up, ..." refer to in your positive prompting? Is this something specific to the LoRA, or perhaps something more general with SD I haven't heard of yet? Thank you!
you should change the title im pretty sure you mean vintage, becomes hard to search for it
LOL never realized I made a typo, changing it right now
Now this is that good shit! - Perfect Cell
Any plans on updating this? It's y favourite lora at the moment!
models aren't like software that need constant updates. it's good as it is.
Great, love it....
Nice lora. I really like the style. I am notice that all images have vignetting. Is there a way to remove that, so colors look more vivid? Or is it a feature of the training set?
just edit it out in lightroom
Seeing all these masterful images, reminds me of the heights of 90s anime. Great work
Any chance I could persuade you to train a Flux lora with your data? :)
ohhhh I totally could! might have to re-tag the data tho so flux understands it
@dalla1re if it helps any, I have been using chatgpt to help write prompts for me based on images I have from other models and it's been working like a dream for flux! Might help expedite the process a bit :p
@cubby_ai Can your share the gpt settings? I will pay with buzz:)
@skelllens962 there's no settings, I just ask gpt to write "a natural language prompt that I can feed to an image generator" and it usually does pretty well with just that.
is it possible you could update it to go lower than 0.5 <3
just did a year later but I stopped AI for a while...
god,can you make a lora of Illustrious?
THIS IS JUST ASTONISHING!!!!!
GREAT
with changes coming to this platform for uk, would love for this model to be implemented on tensor.art so i can continue to create artwork from it at a later date.
id love an illustrious version
we need illustrious version !!
Details
Files
OldAnimeStyle_XL_v3.safetensors
Mirrors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
30-vintage-1990s-anime.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
OldAnimeStyle_XL_v3.safetensors
Vintage_1990s_Anime_SDXL_LoRA_Style_Pony_333041 (V2 17c73ee146) (V3 0252ffccc53f).safetensors
Available On (5 platforms)
Same model published on other platforms. May have additional downloads or version variants.