Finetuning and training takes a lot of time and energy: If you enjoy what I do and want to help, why not donate? Click here!
(Updates and revisions to DREAD will always be free! This is just if you want to!)
A custom build model using SD1.5 as a base. Trained on 30k images ove it's 4 versions, (make that 35000 over 5) of horror and psychological terror, with a very graining and dark aesthetic. The images were hand sorted to make sure they fit the style and theme I was looking for, and after auto-tagging were then manually tagged with more clear identifiers and themes to help push generation even further.
If you are looking for crisp and clean images, go elsewhere. This is one grainy boi.
All sample images are without any post processing, upscaling, or cherry picking. Just random / same seeds as previous versions and posting.
V5 now done! (lora here - is large because when extracting at a lower DIM it wouldn't work well)
Dread-v5 adds more varied images and much more complex tagging for all of them, at a higher resolution. This will be the final Dread version, at least for SD1.5, since there is little else to be done to the base model. From here on out, any additions will be via "expansion pack" lora.
USE TIPS
V5 has many more keywords trained into it than the older versions, so use is both much more fine tuned but also a bit more finicky. The keywords below are ones that can help evoke horror feelings and moods, but are far from necessary:
found footage, disposable camera, trailcam, liminal, abandoned, hallway, house, forest, swamp, vines, roots, branches, dendrophobia, body horror, visceral, circulatory system, viscera, organs, blood, injury, laceration, gore, body horror, boil, rash, peeling flesh, fleshy, wet, water, drowning, burning, fire, flames, injury, plants, mushroom, fungus, diagram, illustration, tome, grimoire, arcane, sigil, magic symbol, experiment, medical, research, lab, malpractice, contained, restrained, chains, cables, cord, torture instrument, dungeon, mansion, manor, estate, swamp, entity, creature, being, monster, animal, glowing eyes, multiple eyes/limbs/mouths/etc, trypophobia, holes, torture, suffering, pain, despair, open mouth, sharp teeth, spikes, religious, demonic, ritual, cult, demonic, angelic, and many more...
Push the style further
Use DREAD in your prompt, all caps, to reinforce the horror aesthetic. Only the best images were tagged with this, and it can change your generations greatly. This is not a required tag, however! All images for training contained the word 'horror' in their tags.
One of the sample images is a grid showing a few prompts of the same seed, but using different samplers- just so you can see how each one influences the output and reading of this model.
Description
Skipping with putting V3 up here because it didn't turn out how I liked and I ended up just going straight into a new system. 20k images trained over 5 epochs, 20 steps each, over V2. Each of those 20k were sorted into tag sets manually so they each had in depth custom horror based tags, before appending blip and git tagging data.
Should be better at manuscripts, darker images, and making things macabre. Keyword not needed but can add to the effect over all: DREAD
FAQ
Comments (18)
How long did it take to train with 20k images over 5epochs 20 steps and on what hardware if I may ask?
Also v4 lora pls. This may aid me in making Nurgle death guard plague marines but I dont wanna download an entire model for that purpose tbh.
Just over 48 hours, locally with a 4090, at a large batch size (that I can't recall...) I had it save training states every epoch and didn't do it all at once, but resumed from the state when I wasn't needing to use my computer.
Also, behold: https://civitai.com/models/47169
I just uploaded the Lora of V4
It is updated? Nice, l loved playing around with older version. One question though, apart from example images, which keywords/terms was it trained on? Stuff like body horror/guro is usually a weak spot for Stable Diffusion (i.e. it is hard to generate a headless horseman and it smooths open wounds since it's trained on... whole... people) and it takes a lot of trial and error on my old PC to get a gritty war scene or a butcher's dungeon.
In the description I added some more tags it was trained on, but there are a LOT not added. If you want wounds, the data set had lots! Tags I suggest that were in the training set: Injury, torture, blood, body horror, macabre, laceration, pain
Good luck, and I am just uploading a LORA of this now so that may also help you in combining it with other models that direct images the way you want.
It's amazing but bro are you alright?
Hah, yes I am fine. I've just grown up with silent hill 1 and from then on loving the aesthetic haha
Amazing model. Love it!
By far my favorite model, I can't stop using it on everything. I prompt DREAD on other models and all they do is disappoint.
Some of the combination images you've posted have been amazing! Thanks for the kind words.
Hey, are you interested in creating a checkpoint mixing dread v4 and my checkpoint terror-yamer-v2?
Hi, I've been eyeing this thing for a while as a possible ingredient in a future merge, i think it could add some moldy zest to the darker side of the cake, given your permissions I've thus far hung my head and walked on. I don't want to restrict my merges so much and if that's a hard line for you I'll give it a miss. But I thought I'd ask.
I haven't started training myself yet, and I get that you want to preserve your work. I'd hate to put so much effort in and for someone to come along, merge it at like a silly small low effort ratio and say "hey, durp, this is mine."
So, if you'd be ok with different permissions, I'd go ahead and start seeing how this and Opiate play together.
Either way, I love what you're doing here!
What sort of merge are you doing? I'm not against my model being used in merges, I just ask it be credited as being a part of said merge. So yeah, go wild! Merge away so long as you credit. The permission thing is more so people ask, like you did. All good- I didn't mean to make it "no comercial selling of images" either, and changed that, thanks for the catch as I completely would have missed that error.
@hobolyra This makes me very happy. I'm working on Opiate V2 atm. You asked a while back what models it was a merge of and I rambled all I could remember at ye. What I'm aiming at is a) making it more trippy, and b) trying to make it more dark|gritty|moody|gruesome, (can we talk to each other like this now?). Also I'm not a fan of super crisp outputs, like the subject is often disconnected from the scene, too clean. And looking at Dread V4 it ticks all of those boxes.
there another models 3 in the cauldron atm and the results are already getting pretty good, already a decent step up from V1, just needs that extra writhing squiggly glue where I can say, "yes, that is a version forward"
Everything I use will be credited added to the recommended resources list (not sure that's the best way, each version will make that list longer, but for now it's a time saver).
Thank you. I'm excited to see what this thing will do, time to download and jump in. Have a good one.
@DominoPrincip Good luck! When you're done and happy with the result, throw me a link to check it out!
As requested: https://civitai.com/models/69587/opiate
Man DREAD V4 really got this one over the line, improvements to be made for sure but the darker sides to this one are... I think I'm developing a monster making addiction. If you give it a whirl I hope you enjoy it. Thank ye kindly.
Every output makes my skin crawl. Incredible job.