The lost, innocent, and very damp wayward traveler from 1975's cult classic musical, "The Rocky Horror Picture Show," Janet Weiss!
The training data for this LoRA is almost exclusively screen grabs from a 50-year-old movie that had a lot of quick motion and often dubious lighting. I did my best to curate quality samples, but certain elements struggled to train properly. In other words, you might have to do a lot of prompt tweaking to get the results you're looking for.
Trigger word: RHPSJanet,
Appearance prompt helpers: blonde_hair, curly_hair, brown_eyes, big_eyes, thin_eyebrows, red_lips, necklace,
Appearance negative prompt helpers: thick_eyebrows, eyebrows, facial_markings, black_hair,
UNDERWEAR: JanetUnderwear, white_bra, white_panties, white_short_skirt,
FLOORSHOW: JanetFloorshow, clown_makeup, white_facepaint, blue_eyeshadow, stage_makeup, mole, strapless_black_corset, front_laced_corset, thighhighs, black_panties, garter_belt, black_elbow_gloves, sequins,
FLOORSHOW negative prompts: blue_facepaint, clown_nose, bra, white_clothing, white_bra, white_gloves, white_panties, hat, forehead_gem, blue_corset,
(With both my Columbia and Janet Weiss models, the training absolutely refused to include their signature lace boas from the floorshow. I tried everything I could think of to drag the boas into images with prompting, but to no avail. There's plenty of references in the training data, so it might be possible to make work, but you're on your own for now. I've made a supplemental Ruffle Lace Boa LoRA that can be found below in the "Suggested Resources" section.)
PINK DRESS: white_hair_ornaments, pink_dress, white_cardigan,
(this wasn't explicitly trained, due to a lack of clean source images, but SOME images made it into the dataset. As such, there is no specific trigger word. Thankfully, the costume is generic enough that most checkpoints shouldn't have much problem filling in the blanks)
***ADVISORY***
This has been flagged as "real person" model, so be careful posting images on civitai that use skimpy outfits or ANY nudity (including cleavage or midriffs); this site's draconian censorship policy has decreed that such content is too revealing or whatever. Chances are that they'll get reported and removed.
Description
Comments (4)
Good model, but perhaps it would
be best to train the model again more pictures of
the real Susan Sarandon actress when she was younger.
If you need a folder with more pictures, you can contact me via Civitai Chat and I can send you a folder with good face images.
If your images of her are not too sharp, or too
small, I can fix most of them using Topaz Pro-Studio and Photoshop. Most images these days can be fixed and even enlarged to a better quality.
Cheers
MAX
Getting quality images of Sarandon as Janet wasn't really the problem with this model; her character actually has the best still-motion closeups of the entire cast. The model works reasonably for what I intended (minus the boa issue, which I intend to fix with a supplemental LoRA), but due to the majority of the character's costumes it is exceptionally horny. This made collecting sample images tricky because ALL OF THEM wanted to be somewhere between underwear-clad and completely nude. Appropriate for the character, but not so much for civitai's policies about depicting 'real' people. Even some of the few close-ups that I submitted got flagged as pornographic because of bare shoulders and visible bra straps.
In retrospect, this entire film probably wasn't a good fit with this site's current policies. But hey, Rocky Horror fans gotta represent, right? ;)
Where I feel this model could use the most work is in the Floorshow makeup department. That whole scene is just so weirdly lit and filmed that I could only scrape together a few decent shots. The dataset is really small, and my prompt-fu wasn't quite up to the task of hammering it into submission. Unfortunately, outside of the film and a handful of production shots, there isn't a lot more to work with.
If I end up re-training this model, I'll probably just grab an HD copy of the film and manually scrape as many screenshots as I can from the first couple of scenes where she still has clothes on, and emphasize the samples around her "Dammit, Janet" and "There's a Light" outfits. In that scenario, I might employ your suggestion, because her facial structure is pretty unusual, and SD struggles with getting her eyes to look accurate.
Thank you for your offer, regardless.
@spiderhaunt Thank you for your response good friend. You are doing a fine job and
I congratule you as a cool artist !
Yes, I am a Rocky Horror Fan since the 1980s and this is such an iconic musical-film which
is loads of fun and has some very memorable characters. Perfect for some cool fantasy
artworks for Stable Diffusion AI Art. Thank you.
...
I will give you some ideas from the days I used to create images of
"celebrity fakes" using Photoshop (I am retired now. lol ).
If you use Photoshop and can get around with doing face swaps and then doing some good
skin color RGB tweaks for everything to match , which is not that difficult, you can always
use your film screen-capture images of her face and place them over an image of a woman
which is fully clothed.
That way, the A.I. will still be able to train itself on images of her face in a fully clothed
female body AND you will not have too many problems with NSFW pictures.
This is possible because I have done such things in the past and also done many tests since 2022
when doing such famous people which only exist in paintings or as Museum WAX figures.
Examples; the maestros Mozart, Beethoven and Bach.
I used pictures from Madame Tussaud´s Wax museum figures, which are faithful representations of
their faces and the A.I. did a really good job at recreating their faces in a LoRa model.
They work really good when you prompt for "a Photorealistic photograph of a ...." or a
"a Realistic photo of a....".
Just an idea for you.
If you ever need my help with something, you can chat with me via Civitai CHAT, and I will
be glad to help out. Its all FREE and I am always happy to help out.
Cheers
MAX
Dammit...
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.