I am admittedly kind of a noob at this, but if you enjoy my humble work, any donations are greatly appreciated: https://ko-fi.com/insult_to_ninjary
Your generosity helps me stay motivated to keep improving and to keep trying new things!
At a Glance:
Base model: Any (tho some work better than others)
Use a VAE?: Yes (baked or your choice)
Clip skip: 1 or 2 (try both!)
LoRA weight: generally from 1 - 0.7 (adjust as necessary)
Hires. fix: use non-default upscaler; denoising strength 0.5 - 0.4
Trigger Words:
text, numbers, graphics - necessary to get the card-like effect
You must use at least one of the above trigger words to trigger the effect, but I strongly recommend using all 3!
You may need to weight the trigger words with numerical values or parentheses if the effect is not strong enough. I generally put all 3 trigger words inside 3 sets of parentheses, like so: (((text, numbers, graphics))).
You will need to remove any common negative prompts that are related to the trigger words. For example, remove words like: text, signature, watermark, username, etc. Some popular negative embeddings also contain these concepts, so keep that in mind when prompting, and don't use any negative embeddings that may interfere with the triggers!
Description:
If you're like me and didn't know what a tart card is, they're essentially cards that advertise certain... "services" (lol)... and are usually tossed all over the ground in popular nightlife areas. The cards don't display any nudity since they're distributed in public. From the ones I've seen, they usually feature a low-quality picture of a nominally attractive woman, combined with some rudimentary graphics, and slapped together into an overall poor display of graphical design. But that's what adds to their charm!
Now, with this LoRA, you can create your own lewd litter in both illustrated and realistic forms! Or, if you prompt less sexily, you can just make some general low-quality-looking cards/advertisements! If you're having trouble using the LoRA, please check the section below.
Troubleshooting:
Using trigger words is necessary! See the section on trigger words above. If you don't use any, you'll just get a crappy picture of a woman. That's no fun without the rest of the package!
That said, some base models simply refuse to produce any text, numbers, or graphics no matter what you prompt. Sadly, you'll just have to give up on those models and try others (unless you can find a solution that I haven't discovered yet!).
The LoRA can struggle with anatomy at times. Lowering the LoRA's weight can correct this, but it also takes away from the style, so it's a double-edged sword. There's always inpainting, or you can just batch gen and hope for the best! I find that clip skip 2 is generally a little better at anatomy, but not always. I've also gotten some nice images at clip skip 1, so play around with both!
The LoRA leans a bit Asian when making realistic images. If you don't like this, you can try to prompt away from it, or use a base model that was trained with more variety.
If you make anything awesome, please consider posting them to the gallery! That goes for all of my LoRAs! I really like seeing what kind of stuff you all are able to make!
If you have any questions, feel free to ask! I will try my best to help! Also, if you have any ideas for future LoRAs, please don't hesitate to share them with me!