⌚ Time traveling detective! Now with all 4 of her outfits and all accessories, as well as a generic "amelia watson" tag for all your other needs. Check the third image for basic prompts and the other images or the right side for extra tags.
Works at weight 1 but might need hires fix sometimes. Made using my Lora guide.
Description
Trained on 450 images of Amelia Watson. Works at weight 1. See trigger words for outfits. Download the dataset if you're interested.
FAQ
Comments (8)
nice boobs, detective
If you need so many tags just to recreate the look, then you’re training wrong imo
You can remove individual items of clothing which you can't do if you hardcode the whole outfit. There's 3 sets of activation tags listed on the side, for each of the 3 outfits, and I don't think any of them are redundant.
Like for sure it's lazier, I'm only pruning tags for hair/eye color, and using a single activation tag instead of separating into folders.
But it's a valid way to do it, I know some very skilled people that are very heavily against pruning ANY tags, which is the opposite of what you want.
Regardless, kind of bold to go to someone's lora and tell them they're doing it wrong without any context.
I literally said imo.
If you don’t agree, you do you.
And no, you can still alter the outfit even with hardcoded ones.
I think it's perfectly fine to leave it unpruned. as it was uploaded to a website that is publicly used by many people, and who's going to know what they might be using it for?
Publishing it with minimal pruning is probably the safest bet for a public use Lora, imo. I know it wasn't related to a training but hey, if the model you're training is flexible and give what you exactly want, it's good enough honestly.
imo making loras like this just adds to their flexibility. the fact that i am able to generate all official outfits just by clicking copy paste on the tags is a lifesaver because it comes out perfect pretty much 100 percent of the time. in addition to all that the lora also has a plain character tag for even more flexibility. if you want to make one lora for a character but you have to spend 30+ minutes to find the exact prompt to match even one of their outfits semi-accurately, then you do you boss.
It's been a couple weeks and I released a new version, I used separate folders and activatation tags like I should have the first time (it's still a learning process for me even though I've written a guide).
However I failed to prune many tags without downsides, so I kept a lot of them just like the first version. This new version took 4 tries to get right partly because of the similar features and many small details. For ease of use I resorted to separating the main tags from the accessories on the tag list. I do think having a single tag per outfit is valid, but it's not my style and I just can't get it to work either.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.