<h1 id="v1(for-anima)-has-been-released.">V1(for <a rel="ugc" href="https://civarchive.com/models/2458426/anima-official">Anima</a>) has been released.</h1>(special thanks to a friend for the help).In this version, I’ve added support for natural language tags, while keeping the original tag system intact.You can now try using natural language prompts when working with the Anima model! (This feature is still in testing.)For the dataset, this release combines both the previous dataset and newly added data (thanks again to everyone who has supported and used this LoRA).Please note that since some AI-generated images are included in the dataset, it may slightly affect facial generation quality. You may need to lower the LoRA strength to achieve better results.<hr /><h1 id="v3(for-illustrious-xl)-:">V3(for <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765/illustrious-xl">Illustrious-XL</a>) :</h1><hr />A friend's commission. I think the effect should be good enough.This LORA should be able to: specify slime color and host body color.For example:<pre><code>goo merging 1, 2boys, red slime, red slime dragon, blue body, blue dog</code></pre>This piece of Prompt is expected to generate a red slime dragon merging with a blue dog.However, it is ok to use (solo) or (2boys) or (duo). This slime has changed a lot, and randomness will give you a lot of interest. Especially in the new version.some AI-generated images (tagged with "goo merging 2"), while the original dataset remains tagged as "goo merging 1"Optional Prompt(Blue font means barely effective,Maybe you can use them for negative prompts):size difference, tentacle in mouth, head back, flexing,head in chest, : (Best used with (size difference), the head of the host will be generated in the middle of slime's chest)<s>ex_merging</s><s>, : (for v2, Tend to generate more chaotic images)</s><s>multiple penis</s><s>, :(for v2, Tend to generate the penis of the host and slime)</s>head in head, : (Like a hooded pajamas, the host head is in slime's head partly)(I really like it, too)tentacle in urethra, slime covered penis, covered penis,Negative Prompt :<pre><code>(sketch:1.2), (monochrome:1.2),(low res:1.2), (watermark), </code></pre>If you want to generate a gray slime, then (monochrome) is unnecessary.<hr /><hr /><h1 id="v2(for-illustrious-xl):">v2(for <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765/illustrious-xl">Illustrious-XL</a>):</h1>This version is trained on the <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765/illustrious-xl">Illustrious-XL</a>. (The number of training steps may be a little small.)And it is recommended to use these model:<a target="_blank" rel="ugc" href="https://civarchive.com/models/827184?modelVersionId=1068947">WAI-NSFW-illustrious-SDXL</a>.It only needs simple quality prompt to generate good pictures. Built-in multiple characters and artist styles, so there is no need to add extra Lora, and the generation speed is fast.Although furry content seems to lack support, it is still a basic model with great potential.<hr /><h1 id="v2(for-indigofurrymixxl):">v2(for <a target="_blank" rel="ugc" href="https://civarchive.com/models/579632?modelVersionId=833339">IndigoFurryMixXL</a>):</h1>Can be used on <a target="_blank" rel="ugc" href="https://civarchive.com/models/579632?modelVersionId=833339">IndigoFurryMixXL</a>.New prompts have been added to help control the relative position(to be verified):(from front position),<ul><li>(mastery position),</li></ul>(from behind position),<ul><li>(mounting),</li><li>(bodyguard position),</li></ul><hr />Although (absorption, corruption, body in body) is mentioned in the title of the model or they are related, these tag is not used in training. I don't know what effect these will have(maybe they are useful). Of course, it will have different effects when used with other lora.Prompt used in the training process:<pre><code>goo merging, tentacles, colored slime, colored body, colored species,</code></pre>After testing, it can be used on a more realistic checkpoint (such as realism). My usage is to generate it with pony v6 first, and then switch the checkpoint.A small issue right now is: how to generate a PG-rated image for display. Haha.<hr />Thank you for reading all the introductions of this Lora and using it to create some goopy things. I made this for fun, too. Welcome and hope to see more images posted below and your comments. I will thumb-up them, and I will pin the images that I think look very very good.<hr />10,000 times! Thank you!

goo merging and absorption [ILXL][Anima][Pony]

{ "engine": "kohya", "unetLR": 0.0005, "clipSkip": 2, "loraType": "lora", "keepTokens": 1, "networkDim": 32, "numRepeats": 2, "resolution": 1024, "lrScheduler": "cosine_with_restarts", "minSnrGamma": 5, "noiseOffset": 0.1, "targetSteps": 1527, "enableBucket": true, "networkAlpha": 16, "optimizerType": "AdamW8Bit", "textEncoderLR": 0.00005, "maxTrainEpochs": 20, "shuffleCaption": true, "trainBatchSize": 6, "flipAugmentation": false, "lrSchedulerNumCycles": 3}

Files with SHA256: 61a6164a278680f7953033ecfdcd5079d26dc90a82a0306f6d025c826bc3f608