Vore Sequence
This LoRa consists in the concept of a vore sequence (as in 'before/after'), as the title suggests.
The trigger keyword is: "vore_sequence", adding the tag "2koma" can help too.
Description
First public version
FAQ
Comments (6)
Hey, love the Lora! I’m often getting blurry or poor-quality faces on this model, but not on your other similar ones. For context, I’m using pretty standard humans, nothing crazy and no other Lora’s. Typically on Prefect Pony. Any advice on improving faces here?
Hi, thanks for the feedback!
The blurred/low res faces are indeed an issue I have been facing when making LoRAs, and I fully intend on iterating on this one to make it better in this regard.
Though, for the time being, factors that I've found to minimize these issues are:
* Lowering the CFG Scale: In general makes the image less glitchy, but follows less of the prompt
* Lowering LoRA strenght: Using a higher strenght seems to make the glitching worse, and I've found some degree of success by lowering it in the resources tab and at the same time increasing the strength of the trigger word by using parenthesis
* Increasing the number of steps (doesn't always work, though): Helps getting rid of smaller artifacts, but is very "hit-or-miss"
* Adding quality LoRAs and embeddings: Things such as easynegative and detail LoRAs have helped to some degree
These are my recommendations for the time being. Some base models work better than others (for example, YiffyMix v60 works great, albeit locally, while Pony Realism tends to be more glitchy), but as you mentioned Prefect Pony, it should work well with it, and the issue is indeed with the LoRA 🙃
I want to try and train this LoRA again (and many others, too), but this time on Civitai's trainer, as I used a different one for this version🤞
Herzlichen Dank!
@oilio2 Thanks for the info, ill give these a shot!
Is a illustrious version planned? I'd love to see it in that quality..
I don't have any experience using illustrious, though I can certainly try!
I already had an update for this one in my bucket list so I don't see a reason to not give this a shot
Finally got around to trying it out, and it turned out surprisingly well when using WAN ckpt, thank you for suggesting it ♥️

















