<h1 id="!!!-download-the-config-yaml-use-a-webui-that-can-do-v-pred-set-cfg-rescale-to-0.7-use-non-karras-samplers-!!!">!!! download the config YAML, use a webui that can do v-pred, set CFG rescale to 0.7, use non-Karras samplers !!!</h1>This is a NSFW-enhanced full finetune of SD2 with v-pred, trained with zero terminal SNR for exceptional light/dark and color reproduction.<ul><li>download the yaml or use the diffusers mirror <a target="_blank" rel="ugc" href="https://huggingface.co/woweenie/pretty-girls-next-door-v9.0">woweenie/pretty-girls-next-door-v9.0</a></li><li>use a webui that supports v-pred and use CFG Rescale=0.7 (InvokeAI with diffusers link recommended because it handles this for you). Scroll down for Automatic1111 install guide.</li><li>use NON-K samplers (ZTSNR does not work with Karras)</li></ul><h2 id="do-not-use-to-make-deepfakes.-do-not-use-to-make-images-of-children.">DO NOT USE TO MAKE DEEPFAKES. DO NOT USE TO MAKE IMAGES OF CHILDREN.</h2>Using an entirely custom dataset, almost all manually captioned. Focus is on softcore and erotica, rather than porn per se. It's really great at doing erotic images of fully-clothed women. It can definitely do nudes, but it's more like z1shy or m3t-4rt than 4tkg4ller14 or "porn" porn. V5 now with all new training methods, highest quality yet.Your prompt should be something like: "a photo of a woman with large breasts and a pretty face. front view, upper body shot. wearing a camisole. 85mm, f/3.5" Scroll down for an indepth prompting instructions.<h1 id="auto1111-install-guide">Auto1111 Install Guide</h1><ol><li>Download BOTH FILES (.safetensors and .yaml, it's under config) files and put them in your webui\models\Stable-diffusion folder.</li><li>Go to the Extensions tab, click Install from URL, and install <a target="_blank" rel="ugc" href="https://github.com/Seshelle/CFG_Rescale_webui">https://github.com/Seshelle/CFG_Rescale_webui</a><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/504a9e3e-1927-4600-b8ac-7ccea485d750/width=525/504a9e3e-1927-4600-b8ac-7ccea485d750.jpeg" /></li><li>Then go back to Installed, make sure CFG_Rescale_webui is on the list at the bottom, and click Apply and restart UI<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a422557e-397b-40ce-9d30-67b3de492f78/width=525/a422557e-397b-40ce-9d30-67b3de492f78.jpeg" /></li></ol>Download the Config too (.yaml file) or you'll get black images. Or use the Diffusers mirror with Invokeai or Comfy: <a target="_blank" rel="ugc" href="https://huggingface.co/woweenie/pretty-girls-next-door-v7.0">woweenie/pretty_girls_next_door_v7.0</a>This model is horny and will produce nudity without prompting. Add "naked" to the negative prompt if you don't want NSFW output.This is a SD2.1 model (with OpenCLIP VIT-L/14 text encoder), so prompting is more like SDXL than SD1.5 - write a mix of partial and full sentences, keep the negative simple. Use the samples as a guide.Recommended settings:<ol><li>DPM++ 2M SDE, 30-50 steps, CFG 4-10, CFG rescale 0.7</li><li>For very high contrast images: DPM++ SDE, 35 steps-50, CFG 4-10, CFG rescale 0.7. The step count has a big effect! More steps seems to make the contrast more intense.</li></ol>Size: Trained on 768 and 1024, you can go much higher in one dimension eg 1050x1427.CFG: generaly 6-10, but you can go lower or higher as long as you use CFG rescale.Recommended negative prompt: as short as possible, start with "corpse, 3d render"<h2 id="prompting-guide">Prompting guide</h2>See below for the full guide it's from v1 but is still mostly correct.You can get big quality imporvement in the normal way, by stating camera settings eg "canon eos" or "analog lens" or "film grain" or "100mm, f/3.5" (experiemnt with different numbers).You can also get a big quality improvement by adding a random person's name, use <a target="_blank" rel="ugc" href="https://www.behindthename.com/random/random.php?gender=f&amp;number=2&amp;sets=5&amp;surname=&amp;randomsurname=yes&amp;all=yes">Behind The Name's random name generator</a> to make up some fake names for you. Changing the nationality of the name changes what the person looks like. They don't have to be real names, I also notice improvemetn just typing "a photo of ufailhk fnlkiha wearing a t-shirt" instead of "a photo of a woman wearing a t-shirt"Pose is the hardest thing to prompt right. this is because it is ambiguous, depending on the camera angle as well as the pose, so you can get better poses by also prompting a view angle eg "front view", "rear view" etc.<ul><li>if the pose still looks like a scrambled combination of multiple poses, try putting the mixed up poses in negative, so let's say you want "lying on her back" and it keeps drawing her butt but her head on backwards then that's because it's confused and trying to also (wrongly) render her "lying on her stomach" at the same time - so put that in negative to stop it.</li><li>this works for other poses too, one pose in positive and one in negative helps to make it solid</li><li>you can also add a view prompt to negative for views you don't want that it is still trying to make</li><li>you can also use "group" in negative prompt or "modernist garbage" which helps to fix the pose sometimes</li></ul>Negative Prompt: try with empty first.<ul><li>if she looks dead use "corpse",</li><li>if she looks like a 3d render use "daz3d" or "3d" or "render",</li><li>if she looks like exaggerated anime proportions or too smooth skin use "anime".</li><li>if she has extra arms or legs, or missing arms or legs use "group".</li><li>if the photo looks like it is from a cheap 1980s camera use "instamatic" "snapshot" "vintage" etc</li><li>if the pose or expression is boring (like a mail-order catalog) try adding some of these: "fashion" "magazine" "advertisement" "catalog"</li><li>if she looks like trash try "hooker" or if the whole picture is gritty try "brushstrokes"</li></ul><ul><li>Jokers: "lumpy", "modernist garbage", "sketch", "draft", "ugly" (but risks sameface), "grimace"</li></ul><ul><li>Sometimes a gen is bad because of a word in negative, and its harder to tell which one so best to keep negative short unless you like to go through and try deleting the words one by one to find the bad one.</li></ul><h2 id="v5-changes">V5 changes</h2><ul><li>better prompts, about 200k steps more training, better training strategy</li></ul><h2 id="v4-changes">V4 changes</h2><ul><li>no need for complicated CFG rescale or custom beta schedule</li><li>pose prompting improvements<ul><li>added "fetal pose" for lying on her back, curled up with her knees at her shoulders</li><li>"supine" is now "lying on her back"</li><li>"prone" is now "lying on her stomach"</li><li>"recumbent" is now "lying on her side"</li><li>full list of pose tags: standing, sitting, kneeling, fetal position, on all fours, squatting, lying on her back, lying on her side, lying on her stomach</li><li>can be combined with: closed legs, parted legs, spread legs, crossed legs, one leg raised, both legs raised</li></ul></li><li>better view control<ul><li>3/4 view is now "angled view", you can use "angled front view" or "angled rear view" ensure front/back view</li><li>add "from below", "from above"</li><li>full list of view tags: front view, angled front view, side view, angled rear view, rear view, from above, from below, high angle, low angle</li></ul></li><li>improvements to wet skin prompts (try "wet skin", "soapy skin", "wet see-through t-shirt")</li><li>full list of trained clothing tags, bold is best: bikini, bikini bottoms, bikini top, blanket, bodysuit, bra, camisole, coat, crop top, crossover top, dress, garters, glasses, goggles, gown, gym shorts, halter top, handbra, hat, hoodie, jacket, jeans, jumper, lace-up top, leggings, lingerie, long pants, long sleeved top, nightgown, off-shoulder top, one piece swimsuit, overalls, pajamas, panties, pantyhose, piece of fabric, robe, scarf, shirt, shorts, skirt, sleeveless top, slip, sneakers, sock, socks, sports bra, sunglasses, sunglasses on her head, sweatpants, t-shirt, tank top, tied shirt, top, undersized bikini, wetsuit</li></ul><h2 id="v3-changes">V3 changes</h2><ul><li>ZTSNR training, needs trained beta</li></ul>V3 is only available as diffusers -v3 <a target="_blank" rel="ugc" href="https://huggingface.co/woweenie/pretty_girls_next_door_v3">woweenie/pretty_girls_next_door_v3</a>v3.1 <a target="_blank" rel="ugc" href="https://huggingface.co/woweenie/pretty_girls_next_door_v3.1">woweenie/pretty_girls_next_door_v3.1</a>Trained on <s>4500</s> 12000 images (v1 dataset attached to download) from well-known soft porn websites, and a low alpha merge with Artius-v21 and Mangled_Merge_v3. Mix of manual captions and LLAVA autocaptions with manual tag guidance. Focus is on natural language prompting and flexibility and precision with clothing and nudity.All of the samples are made with diffusers. Generation only, no inpainting or img2img. Seeds are CPU ("random number generator source" option in automatic1111 or "generator source" option in vladmandic).<h2 id="v2-changes">V2 changes</h2><ul><li>better response to shorter prompts</li><li>general image quality improvement</li></ul><h1 id="usage-guide">Usage guide</h1>Download the .yaml or you'll get black images. If you want to reproduce the samples use InvokeAI or vladmandic with diffusers backend.Recommended negative3d render, ugly, misshapen, corpse, brushstrokes, child<h3 id="recommended-settings-(v2)">Recommended settings (V2)</h3><ul><li>DPM++ SDE Karras, CFG 6-8, 24-36 steps</li><li>DPM++ 2M SDE Karras, CFG 6-8, 24-48 steps</li><li>DDIM or Euler a, CFG 6-11, 48-96 steps</li></ul>Recommended settings (V1)<ul><li>DPM++ 2M SDE Karras, CFG 6, 24 steps or 30 steps</li><li>DPM++ SDE Karras, CFG 6, 24 steps or 30 steps</li><li>DDIM or Euler a, CFG 6-11, 48-96 steps</li></ul><h1 id="how-to-prompt">How to Prompt</h1>The model was trained on multiple sentences separated by full stops, and if you prompt like that you will get more out of it. Eg "a photograph of a woman on the beach, wearing a bikini. beautiful soft rim light, intricate details, 8k."<h3 id="prompting-walkthrough">Prompting walkthrough</h3>Start simple. Don't use tag lists, just write something short like "photo of a naked woman at the beach":<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/95f8651b-d5ef-48aa-b9bc-6a7ff4cf5914/width=525/95f8651b-d5ef-48aa-b9bc-6a7ff4cf5914.jpeg" />To get closer, add more words for skin or body parts, but this can make generation weird:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/76d93b9a-1206-46a7-ab3d-7bd12d44f0af/width=525/76d93b9a-1206-46a7-ab3d-7bd12d44f0af.jpeg" />To fix weirdness add a pose to the prompt:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/629ec77c-5fce-4275-99d8-2cea3b84e41b/width=525/629ec77c-5fce-4275-99d8-2cea3b84e41b.jpeg" />Bring the face back into frame by prompting it:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/84d9aa99-f139-425f-b580-e1a04963127a/width=525/84d9aa99-f139-425f-b580-e1a04963127a.jpeg" /><h3 id="to-move-further-back-add-background-details:">To move further back, add background details:</h3><h3><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fb79df52-279b-48ba-bf4f-3df7aaa967bb/width=525/fb79df52-279b-48ba-bf4f-3df7aaa967bb.jpeg" /></h3>You an also change the influence of background details by moving them to a different place in the prompt. If they're closer to the start they will have more influence.<h3 id="more-tips">More tips</h3><ul><li>Increase quality with words like "ultra realistic, skin details, skin pores, 8k, intricate details".</li><li>"Detailed face" or just "face" improves faces, "eyes" improves eyes, "lips" improves mouth shape.<ul><li>"Painted eyes" and "cartoon face" in negative help if the eyes look like they're painted on</li></ul></li><li>Long negatives are often worse. Start short and add words as bad things appear. Clean up words that are doing nothing as you go.</li><li>Prompt a pose to improve results (try to match what the random seed is drawing on its own)Prompt leg pose to improve results (try to match what the random seed is drawing on its own)</li><li>Sometimes "professional photograph" "stunning photograph" "detailed photograph" help, sometimes they make the image worse.</li><li>Sometimes the model gets too close up. Use negative prompt "close up" to fix</li><li>Some things are overtrained in the base model, especially "bedroom", so move it further down the prompt or downweight it with [[bedroom]]</li><li>Adding words like "face" or "breasts" or "hips" or "knees" brings camera closer to those things (or further out of they're not in frame already)</li><li>The first word has a large influence on results. Try "HQ photo", "ultra realistic photo", "professional photo", "digital photo", "detailed photo", "stunning photo", and just "photo"</li><li>Adding camera, lens, or film words eg "Canon EOS Rebel, f/11 35mm" or "ektachrome ER film" sometimes helps, but sometimes make the output less nude</li><li>If some things are dominating, change the order of words.</li><li>Sheer/wet can benefit from "sheer clothing", "see-through", "wet clothing", "visible nipples"</li><li>Try "open" clothing, eg "open shirt", "open robe". Also try "unbuttoned" and "unzipped" - "wearing an unzipped one piece swimsuit" is fun.</li><li>Increase cleavage with "cleavage" or a longer phrase like "a tank top that shows her cleavage"</li><li>To fix cut-off legs or arms put "stump" or "amputee" in negative, or "weird anatomy" or "strange anatomy" or "bad anatomy" </li><li>"Good anatomy", "beautiful composition" sometimes helps</li></ul><h1 id="trained-words">Trained words</h1><h3 id="face">Face</h3>pretty face, cute face, beautiful face, ugly face (for negative)<h3 id="pose">Pose</h3>reliable: standing, kneeling, squatting ok: on all fours, sitting unreliable: prone (lying on stomach), supine (lying on back), recumbent (lying on side)<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9cb6cef6-b6f5-430a-b778-54a70388648f/width=525/9cb6cef6-b6f5-430a-b778-54a70388648f.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/8e80de1e-251f-4755-a196-572b16fc6f2a/width=525/8e80de1e-251f-4755-a196-572b16fc6f2a.jpeg" /><h3 id="body:">Body: </h3>large breasts, medium breasts, small breasts, cleavage, cute butt, round butt, broad hips<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6d089fc6-8d95-43eb-96b9-3fcb1cccc99a/width=525/6d089fc6-8d95-43eb-96b9-3fcb1cccc99a.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4c227556-ecfb-421f-b894-453b33a3bfaa/width=525/4c227556-ecfb-421f-b894-453b33a3bfaa.jpeg" /><h3 id="view:">View:</h3>front view, 3/4 view, side view, rear 3/4 view, rear view, low angle, high angle, overhead shot, medium shot<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1a432ea4-a192-4182-956b-7024ab899171/width=525/1a432ea4-a192-4182-956b-7024ab899171.jpeg" /><h3 id="scene:-(also-good-for-negative)">Scene: (also good for negative)</h3>At the pool, at the beach, at a river, in the shower, in the bath, in a photo studio<h3 id="legs-pose-(can-be-combined-with-pose-doesn't-always-work):">Legs pose (can be combined with pose, doesn't always work): </h3>legs together, legs apart, legs spread, legs crossed<h3 id="clothing-(some-work-better-than-others):">Clothing (some work better than others):</h3>Whole body: dress, overcoat, robe, bathrobe, pajamas Top: tank top, t-shirt, shirt, jumper, sports bra, jacket, hoodie, halter top, polo shirt Bottom: leggings, jeans, skirt, shorts, sarong, socks, sneakers Swimwear: bikini, bikini top, bikini bottoms, one piece swimsuit, wetsuit Underwear: bodysuit, bra, panties, nightgown, slip Face: glasses, sunglasses, hat<h3 id="expressions-(unreliable):">Expressions (unreliable):</h3>duchenne smile, friendly smile, flirty smile, fake smile, polite smile, laughing, confident, distracted, aroused, annoyed, embarrassed, bored, scared, dominating<h3 id="clothing-state:">Clothing state:</h3>Wet, mesh, sheer (combine for better effect eg "wet mesh sheer see-through t-shirt"), fishnet, thin, open, unbuttoned, unzipped, undressing, unbuttoning, unzipping<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/97f8ab74-bcd0-4a25-857a-2bfa5f4002a9/width=525/97f8ab74-bcd0-4a25-857a-2bfa5f4002a9.jpeg" /><h3 id="pussy-(can-be-combined-like-&quot;hairy-puffy-pussy&quot;):">Pussy (can be combined like "hairy puffy pussy"):</h3>Hairy pussy, shaved pussy, bald pussy, tulip pussy, puffy pussy, curtains pussy, horseshoe pussy<h3 id="extras:">Extras:</h3>Wet skin, with her arms raised, with her hand on her hip, nipple pokies (unreliable), strap slipping off her shoulder, lifting her shirt, hiding her pussy between her legs<h1 id="updates">Updates</h1>I have updates to the data source planned when I can find some time to work onthem:<ul><li>Fix prone + front view + low angle + large breasts</li><li>Add sports bra rear view</li><li>Fix nipple pokies?</li><li>Fix undressing?</li><li>Dress, long with legs not visible (leg poses under dress)</li><li>Sheer sports bra?</li><li>Dress/gown rear view</li><li>Check for "legs crossed" and "legs together" overlap, and "legs together" "legs apart" overlap</li><li>Add bikini + small breasts</li><li>Add bodysuit + rear view</li><li>Why is "slip" so weirdly close-up?</li><li>Sheer + naked? It seems like leaving off "naked" means "sheer" can't draw nipples</li><li>Remove natural light?</li><li>More wetsuits</li><li>Add half-ponytail hair</li></ul>For now the dataset is attached, please credit me if you use it.

Pretty Girls Next Door

refined, new training words (visible nipple outline instead of covered nipples, lying face-up/lying face-down instead of prone/supine)

prettyGirlsNextDoor_v50.yaml

prettyGirlsNextDoor_v50.safetensors

v5.0

v9.0

v8.0

v7.0

v6.0

v5.1

v4.0

v2.0

v1.0

complete overhaul, very many Tricks used. 22,000 manually tagged Images mainly from z1shy, m3t4rt, f3mj0y, 4m0ur4ng3ls and mp1stud10s, 10,000 AI tagged Film grabs from <a target="_blank" rel="ugc" href="http://shot.cafe">shot.cafe</a>, some AI Gens scraped from Unstable Diffusion Discord.if you are on AUTOMATIC1111 !!! USE THE CONFIG FILE !!!recommended settings: use a stochastic sampler (SDE or Ancestral)DPM++ SDE, 30 steps, CFG 7-10DPM++ 2M SDE, 50 steps, CFG 7-10negative Prompt: Gens are better if it is short. start with "corpse, ugly", some extra Words i use to help fix bad Gens are "anime" or "render" if it looks fake, use "group" or "modernist garbage" to fix too many Legs Problem!!! DO NOT USE TO MAKE DEEPFAKES OF REAL PEOPLE !!! !!! DO NOT USE TO MAKE PICTURES OF CHILDREN !!!Mostly the same as Earlier Version,you can try these Words, some are working better than the others<pre><code>small breasts, medium breasts, large breasts, huge breasts </code></pre><pre><code>puffy pussy, tulip pussy, curtains pussy, hairy pussy</code></pre><pre><code>closed legs, parted legs, open legs, crossed legs</code></pre><pre><code>front view, (front) angled view, side view, angled rear view, rear view, high angle, low angle, from above, from below</code></pre><pre><code>high contrast, low contrast</code></pre><pre><code>warm color, cool color</code></pre><pre><code>close up, upper body shot, medium shot, wide shot</code></pre>there are very many Clothing Words and Modifiers, you can try<pre><code>wearing a wet see-through t-shirt
wearing a see-through jumper
wearing a thin camisole
wearing a loose hoodie
wearing a blue cotton dress</code></pre>

prettyGirlsNextDoor_v40.safetensors

prettyGirlsNextDoor_v40.yaml

<h1 id="limitations">Limitations</h1>despite have a relatively big data set for my training, the descriptions are extremely simple and are for the most part only the name of the texture itself. That means that the AI doesn't know what a lot of things are supposed to look like.For example, it knows what a sword should look like and understands that a Diamond Sword is blue, and a Ruby Sword is red, but it doesn’t know what the handle or the blade of a sword is, so if your prompt is something too specific, it will probably give you a crappy result.To counter this problem, I will manually edit the descriptions of the 3000+ textures as well as remove the weird and redundant textures that could confuse the AI. If you want to contribute and help me on this project, you are more than welcome. There is no need for new textures at the moment, adding good descriptions is what the training data is lacking.<h1 id="contributions">Contributions</h1>If you wish to contribute to the project, you can directly edit the descriptions of the different items located in the "image/5_minecraft" folder. When modifying the text file, please be as precise as possible in describing the item. Here's an example for a diamond sword: "Diamond sword, Diamond blade, Diamond Hilt with a wooden grip." Something like that should do the trick. Although I could have explicitly mentioned a blue blade in the description, in Minecraft, diamond items are generally associated with the color blue/turquoise. By omitting that detail, the AI can infer that diamond items should be blue. If you wish to avoid AI inferences, you need to add complexity to the description.When creating descriptions, the key objective is to ensure that the AI understands the appearance, nature, and specific characteristics of a given item.<h1 id="to-the-authors-of-the-different-assets">To the Authors of the different assets</h1>If you are the author of certain assets used in training the AI and would like them to be removed, please reach out to me via Discord (Jack Bagel#5043) or open an issue specifying your request to have your assets removed from the dataset. I will promptly delete any models that utilized your assets and create new ones accordingly.<h1 id="credits">Credits</h1>Here is a list of all the mods and assets I used for training this model, along with their respective licenses:Druidcraft (GNU Lesser General Public License)SpellBladeNext/Fabric (MIT)Artifacts (MIT)RLArtifacts (MIT)Better End (MIT)Better Nether (MIT)Farmer's Delight (MIT)Enderscape (MIT)Quark (CC BY-NC-SA 3.0)TwilightForest (CC BY-NC-SA 4.0)Nether's Delight (MIT)Ice and Fire: Dragons (GNU Lesser General Public License)Reliquary Reincarnations (GNU Lesser General Public License)Alex's Mobs (GNU Lesser General Public License)Botania (Botania custom license)Ender IO (Public Domain based)Immersive Armors (GNU Lesser General Public License)End Remastered (I'm the owner and give permission to anyone to train on my mod assets)Overite (MIT)Echoing Depths (MIT)malcolmriley (Creative Commons Attribution 4.0 International)If I have made any mistakes regarding the licensing of these mods, please let me know.

Minecraft Texture Creator

Minecraft-Textures-LoRA-V1.0.safetensors

Minecraft-Textures-Properties-V1.0.yaml

Jack-Bagel MC Textures

A versatile model for creating icon art for computer games that works in multiple genres and at various levels of detail.I'm happy with where this is, so there won't be further updates until there's a new major version of Stable Diffusion.

Stylized RPG game icons

VERY IMPORTANT: FOR THIS TO WORK WITH AUTOMATIC1111, BE SURE TO ALSO DOWNLOAD THE CONFIG FILE AND PLACE IT IN YOUR MODELS FOLDER ALONGSIDE THE MODEL. This may change in the future as automatic1111 support for 2.x matures.This is a finetune of Stable Diffusion 2.1 (as opposed to 1.4), and the results it produces are way more interesting. Click the 'i' icon on the example images of how to prompt it (negative prompt is important here). If you feel like it, please experiment and leave any interesting discoveries in the reviews.License: I don't believe models ought to be copyrightable, so you can do whatever you want with this, including uploading and redistributing it elsewhere, finetuning it, berrymixing it, etc. I do humbly request that you share any models that you make with this one, but you're free not to.This model was finetuned with <a href="https://github.com/devilismyfriend/StableTuner" rel="ugc" target="_blank">StableTuner</a> with a dataset of ~250 manually captioned images. Learning rate was 1e-6, 250 epochs, batch size of 24. Text learning cutoff was 75 epochs to prevent overfitting.Note: I've accidentally discovered that this model does some amazing non-icon art. See the cityscape picture for an example of another way to prompt it.

stylizedRPGGameIcons_v4SD21.ckpt

stylizedRPGGameIcons_v4SD21.yaml

v4 SD 2.1

Trained upon <a target="_blank" rel="ugc" href="https://github.com/microsoft/fluentui-emoji">Microsoft Fluent-Emoji</a> which is described as "a collection of familiar, friendly, and modern emoji from Microsoft"The Emoji file names were converted to become the text descriptions. It made the model learn a few special words: "flat", "high contrast" and "color"

FluentUI Emoji

This is a v2.1 derived version. You will need to copy both the .ckpt as well as the .yaml file to your models/Stable-Diffusion folderStable-Diffusion v2.1 fine-tuned for 10k steps using <a target="_blank" rel="ugc" href="https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image.py">Huggingface Diffusers train_text_to_image script</a> upon <a target="_blank" rel="ugc" href="https://huggingface.co/datasets/Norod78/microsoft-fluentui-emoji-512-whitebg">Norod78/microsoft-fluentui-emoji-512-whitebg</a>This model version is also available in both Diffusers as well as Checkpoint formats on its <a target="_blank" rel="ugc" href="https://huggingface.co/Norod78/sd21-fluentui-emoji">Model card in Huggingface</a>The inference widget embedded in the model card can be used for quickly trying it out

fluentuiEmoji_sd21.ckpt

fluentuiEmoji_sd21.yaml

sd21

sd15

<ul><li>download the YAML </li><li>use a webui that supports v-pred, turn on ZTSNR betas, and use CFG Rescale=0.7 (InvokeAI with diffusers link recommended because it handles this for you</li><li>use NON-K samplers (ZTSNR does not work with Karras)</li><li></li></ul><h3 id="if-you-don't-you-will-get-blackbrokenweird-images-8n20ubx84">if you don't YOU WILL GET BLACK/BROKEN/WEIRD IMAGES</h3>Photography model, resumed from <a target="_blank" rel="ugc" href="https://civarchive.com/models/120851">Pretty Girls Next Door </a>with new training data - prompting tips can be found <a target="_blank" rel="ugc" href="https://civarchive.com/models/120851">here</a>. The model uses an OpenClip-H text encoder. The prompting style is halfway between SD1.5 and SDXL. Some of the training data is booru tags, some is moondream English sentences, some is hand-labelled.Recommended settingsThis is a v-pred model trained with ZTSNR. Use CFG Rescale = 0.7 and enable zero-terminal SNR betas. For best results, use InvokeAI.Image size; 640x896Steps; 50Scheduler; DPM++ 3M or DPM++ 2M SDE (NOT KARRAS)

horsy

horsy_v10.safetensors

horsy_v10.yaml

<h1>D&amp;D top down token</h1><h2>Resolution</h2>Model was trained on 512x512 images, so it works best on this resolution.<h2>Prompting</h2>Model was trained on D&amp;D monsters, heroes and master characters (NPCs). So it works best for medieval fantasy setting. It's recommended to use trigger words, that define monster type.But it's not overtrained only on this setting, as for me works pretty good with almost everything I tried. You can not use trigger words at all, that would be fine too, sometimes that's even better.Also you can try mix them to get interesting results. Check all example images for inspiration.Link to huggingface: <a target="_blank" rel="ugc" href="https://huggingface.co/Zapper/top-down-token-v1-0">https://huggingface.co/Zapper/top-down-token-v1-0</a>

D&D top down token

dDTopDownToken_v10.yaml

dDTopDownToken_v10.ckpt

<h3>Stable-Diffusion Hearthstone card generator</h3>You can <a target="_blank" rel="ugc" href="https://huggingface.co/datasets/Norod78/hearthstone-cards-512">browse the training dataset</a> to understand how the model learned its text prompt and try to mimic it.First digit in prompt controls the Mana-cost (pretty well)The digit can be followed by card name, then special ability and description, a "." and space.The prompt should end with the text "Hearthstone card".<h2>Please note that the entire training set contains actual Hearthstone card images which are copyrighted by Blizzard </h2>So it is possible that the generated images contain copyrighted elements and should only be use for your private entertainment

Hearthstone cards

Full SafeTensors version of the Stable-Diffusion v2.1 fine-tuned for 10k steps using <a target="_blank" rel="ugc" href="https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image.py">Huggingface Diffusers train_text_to_image script</a> upon <a target="_blank" rel="ugc" href="https://huggingface.co/datasets/Norod78/hearthstone-cards-512">Norod78/hearthstone-cards-512</a>

hearthstoneCards_v21.yaml

hearthstoneCards_v21.safetensors

v2.1

<h1 id="heading-1682"><a target="_blank" rel="ugc" href="http://di.FFUSION.ai">di.FFUSION.ai</a>-tXe-FXAA</h1><h3 id="heading-1683">Trained on "121361" images. <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ad54f8db-530b-4d20-97c0-c546c0d7988e/width=525/ad54f8db-530b-4d20-97c0-c546c0d7988e.jpeg" /> </h3>Enhance your model's quality and sharpness using your own pre-trained Unet. The text encoder (without UNET) is wrapped in LyCORIS. Optimizer: torch.optim.adamw.AdamW(weight_decay=0.01, betas=(0.9, 0.99))Network dimension/rank: 768.0 Alpha: 768.0 Module: lycoris.kohya {'conv_dim': '256', 'conv_alpha': '256', 'algo': 'loha'}Large size due to Lyco CONV 256<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/58fb1eb8-2eaf-416d-9511-80c996f9ecc1/width=525/58fb1eb8-2eaf-416d-9511-80c996f9ecc1.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/2858847c-3f55-48c5-b127-472716acaafc/width=525/2858847c-3f55-48c5-b127-472716acaafc.jpeg" />For a1111 Install <a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris">https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris</a>Download <a target="_blank" rel="ugc" href="http://di.FFUSION.ai">di.FFUSION.ai</a>-tXe-FXAA to /models/LycorisOption1:Insert &lt;lyco:<a target="_blank" rel="ugc" href="http://di.FFUSION.ai">di.FFUSION.ai</a>-tXe-FXAA:1.0&gt; to prompt No need to split Unet and Text Enc as its only TX encoder there.You can go up to 2x weightsOption2: If you need it always ON (ex run a batch from txt file) then you can go to settings / Quicksettings listadd sd_lyco<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/cdf94aee-f8c9-4d9a-a3e9-ed2e6cdbb955/width=525/cdf94aee-f8c9-4d9a-a3e9-ed2e6cdbb955.jpeg" />restart and you should have a drop-down now 🤟 🥃 <img src="https://user-images.githubusercontent.com/50985923/238659318-db7ea36c-4d59-4c9f-83c0-e0cc09741e64.png" alt="image" />More info: "ss_text_encoder_lr": "1e-07", "ss_keep_tokens": "3", "ss_network_args": { "conv_dim": "256", "conv_alpha": "256", "algo": "loha" }, "img_count": 121361 } "ss_total_batch_size": "100", "ss_network_dim": "768", "ss_max_bucket_reso": "1024", "ss_network_alpha": "768.0", "ss_steps": "2444", "sshs_legacy_hash": "539b2745", "ss_batch_size_per_device": "20", "ss_max_train_steps": "2444", "ss_network_module": "lycoris.kohya", This is a heavy experimental version we used to test even with sloppy captions (quick WD tags and terrible clip), yet the results were satisfying.Note: This is not the text encoder used in the official FFUSION AI model.

di.FFUSION.ai Text Encoder - SD 2.1 LyCORIS

di.FFUSION.ai-tXe-FXAA.safetensors

di.FFUSION.ai-tXe-FXAA.yaml

FXAA v1.0

A dreambooth-method finetune of stable diffusion that will output cool looking robots when prompted.UsageKeep the words <code>nousr robot</code> towards the beginning of your prompt to invoke the finetuned style. Use negative prompts to achieve the best result.Created by <a target="_blank" rel="ugc" href="https://huggingface.co/nousr">nousr</a>.source: <a target="_blank" rel="ugc" href="https://huggingface.co/nousr/robo-diffusion-2-base">https://huggingface.co/nousr/robo-diffusion-2-base</a>

Files with SHA256: ec4094bd36ef558cd9d52c9985471ac91d2c0c5bd071717411a4ccd5adf4f311