F@rt model

<p>QUICK INFO:Trained on NAITrained on 400+ images at 768x768Recommended Weight: 0.7 (Feel free to experiment. When combining with other LoRAs you may get a better quality image by lowering the weight to 0.6 but this might make generating the fart effect harder.)Activation Tags: Fart, Farting, Yellow GasUseful-ish Tags: Facefarting, Sittingfart, Povfart, Frontfart, Straining, Bloated BellyGUIDE: Click here!IMPORTANT INFO:I tried my hand at making a Fart LoRA, I'm relatively new to making LoRAs but I've done the best I can. I started working on this before the other two Fart LoRAs on this site were uploaded. While I named this LoRA the "Ultimate Fart LoRA" the truth is that this LoRA is a little less "ultimate" than I initially intended and a lot of things may need some trial and error (luck) to pull off.Like with messy fingers, the gas effects have a tendency to be messy too. I really do believe this is just AI being AI and I don't know what can be done to make it more consistent. The AI does not understand what a fart cloud is, so it is common for it to mess up. You just gotta keep generating until the AI gets it right. If you've been working with AI art for a while you're probably very familiar with that already anyways. Some Checkpoints/LoRAs work well with this LoRA and some Checkpoints/LoRAs don't work well with this LoRA. I would assume that this LoRA works best with detailed anime-style models. How well it works with other styles will depend on how well that model works with this LoRA. The showcase images here were made using Kotosmix which has worked pretty well for me thus far. As I said, other models are also capable but the quality of the image will depend on which model you use.This model will not generate onomatopoeia and other effects such as rumbling bellies which may be commonly associated with fart images. It is better if you edit that in with your editing software of choice if you desire it. It doesn't take much effort to do and it's better that way since including such images in the dataset may negatively effect the LoRA.Lastly, this model has a habit of removing clothing that covers a characters ass for some reason. I don't know if it's because there's a large bias of naked or pantless girls in the dataset or if the AI is somehow confusing the gas effects for skin. It doesn't always happen but if it does you may want to up the priority on certain clothing or whatever else to counter this.TAG INFO:Tagging is split into mandatory and experimental tags. Mandatory ones are the ones you're gonna be needing to use all the time and experimental tags can be used depending on the scenario however some of them don't really work. Some of the experimental tags are necessary because fart gas effects change shape and design depending on the camera angle and scenario.MANDATORY TAGS:Fart, Farting - These two tags activate the LoRA and are required.Yellow Gas - This is not mandatory but recommended. It seems this tag has an unknown effect on the gas effect which I haven't fully tested. I myself always include this tag.USEFUL-ISH TAGS:Facefarting - For when a character is farting on another character, typically in their face. This tag can be helpful and is recommended to use if you are generating facefart content.Sittingfart - For when a character is farting while sitting down on something, except for when sitting on someone's face in which the facefarting tag is more appropriate. You can use both but... I don't know if that makes it better or worse. This tag seems to be helpful, but it's difficult to gauge how effective it is.Povfart - To be used when a character's ass is directly facing the viewer and the character is not at an angle to the camera. Do not use this in normal "from behind" scenarios, only when the character's ass is directly facing the viewer or in close-up images. It seems a help... a little bit... maybe... I'd say God knows but he's abandoned me at this point.Frontfart - To be used when the character is directly facing the viewer and is not at an angle. If the character is at any sort of angle, do not use this. This tag might not work anyways.Straining - This tag assists in a certain facial expression where the character is pushing hard to get a fart out and is in embarrassed discomfort. Use alongside tags such as clenched teeth, wince, pain, embarrassed, v-shaped eyebrows, nose blush, etc.Bloated Belly - This tag assists in giving a character a big, usually round, belly which is a common way of showing a character is full of gas. Use this tag alongside the big belly or huge belly tags (huge belly is probably better).EXPERIMENTAL TAGS: (NOT RECOMMENDED)Gasstream - When the gas effect comes out of the character's anus in a long thin stream before spreading out. This tag may not be necessary as the AI has the tendency to do a gas stream on its own even without including this tag. It may help force it if you really need it but I ultimately recommend not using this tag.Smallfart, mediumfart, bigfart - These tags control the size of the fart. I don't think they work. I never use them in my generations.Simplegas - Tagged onto images where the gas effects are very simple, such as only being one color. This tag may have been more useful if there were more simple gas effect images in the dataset but there's only a small minority of them which ultimately makes this tag unnecessary to use.Sound effects, onomatopoeia - The vast majority of images in the dataset don't have text and onomatopoeia in them which means it is usually exceptionally rare for them to show up during image generation. For that reason, adding these tags is unnecessary and it may do more harm than good. If you want images to actually have onomatopoeia you are much better off opening your editing software of choice and adding them yourself. Like with simplegas, it would've been more useful if there were more images in the dataset with onomatopoeia. Do not use.Unorthodoxgas - An attempt to prevent messy gas effects by tagging the images where the gas effects are wonky. This tag does not work.Green gas, Brown gas, Orange gas</p>

Files with SHA256: c5dbd27741b1507e22abf4571588c99a66fde3168413293d501bc3ac4d69b571