This is my new Model! If you want your own style/person trained, just DM me with a dataset of ~20 imgs on discord Flonix#4022 and ill train it for you!
My first Lora made on my Danmumford Dataset, set the weight lower than 1 to get it to follow your prompt more, no Tag needed. Also remember to set CLIP skip to 2.
Keep in mind this is LoRA, so you need to use it as such. It's not a standalone model and it's not a TI.
may do accidental nsfw
How to use LoRA's in auto1111:
Update webui (use
git pulllike here or redownload it)Copy the file to
stable-diffusion-webui/models/loraSelect your LoRA like in this video
Make sure to change the weight (by default it's
:1which is usually too high)
usage by @Lykon
Description
FAQ
Comments (9)
Does this model suffer from the 'similar faces' problem? If not, what method do you use to make the training not fall into that rut?
not sure what you mean with the similar faces problem?
@Floniixcorn When training style it tends to generate a merge of the faces that were present in the training image set, as then there is a hard inclination of the model to generate faces that look like it.
@slashkey i dont know if this model has it, but i dont think so, it wasnt trained on faces
@Floniixcorn Ah okay... Maybe styles should exclude all faces. There is a lora training repo that has the ability to reduce the impact of faces on style. https://replicate.com/cloneofsimo/lora-advanced-training I haven't tried it yet because I prefer to run locally.
@slashkey i think you could fix the faces problem by getting more different faces in your dataset and using accurate filewords so that the ai knows what it learns
Is CLIP skip necessary? I heard it was just placebo and only works on models made from invokeAI...
not sure, but ive gotten good results with it on
You can try generating with the same seed value with CLIP skip set to 1 and 2 and comparing.
My results look better with CLIP skip = 2, but then again I used Anything-V4.5. Your model might work better with it set to 1. It certainly has a visible effect
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
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.