<h3 id="heading-1023">一、模型概述：</h3>时至今日，我们依然在怀念艾泽拉斯大陆；依稀记得还是在10年前，我选择了亡灵法师，开启了魔兽世界之旅；曾几何时，为了完成一系列成就而废寝忘食，为了组队开团下副本定时调闹钟，为了抓一个宝宝坐骑，彻夜蹲守......训练此模型的初衷，是为了纪念逝去的青葱岁月，所有训练素材来源于网络图片，游戏截图等；你可以尝试用此模型生成图片作为壁纸，偶尔翻翻看看，以当纪念你的魔兽世界。<h3 id="heading-1024">二、模型使用说明：</h3>1、推荐lora权重：0.7-0.9，添加触发词：wow2wow，2、推荐加载基底大模型：<a target="_blank" rel="ugc" href="https://civarchive.com/models/7371?modelVersionId=43978">ReV Animated_v1.2.2</a>，3、迭代步数：25-40，CFG scale：10-15，Clip skip：2，4、推荐采样器：DPM++ SDE Karras，5、VAE： <a target="_blank" rel="ugc" href="https://huggingface.co/mmmoof1/vae-ft-mse-840000-ema-pruned/blob/main/vae-ft-mse-840000-ema-pruned.safetensors">vae-ft-mse-840000-ema-pruned</a>，6、对于分辨率的设置，建议横版：768x512（或960x768），竖版手机尺寸：512x768（或768x960），得到以上图片之后，可后续使用高清修复（重绘幅度0.3-0.4，放大算法R-ESRGAN 4x+）放大2倍，以增加丰富图片细节；当然，你也可以选择不用高清修复，使用Tiled Diffusion + Tiled VAE放大手段，那样会得到更加不错的高质量图片，7、这个模型的泛化性还行，可以在不同底模下出图；比如切换到真实风格或动漫风格大模型下生成一些有趣的图片，小伙伴们可以尝试。<h3 id="heading-1025">三、其他说明：</h3>1、魔兽lora标准简短调用词：wow2wow,&lt;lora:wow2wow:0.9&gt;,darkness,star \(sky\), weapon,glowing,masterpiece, best quality, most detailed, best aesthetics, highest contrast, highest resolution, best details, 8K wallpapers,直接加载lora，添加触发词：wow2wow，再配合一些提示词即可得到一些关于魔兽世界风格类型的场景和人物图片，2、推荐的一些其他提示词：darkness,weapon,star\(sky\),glowing,android,robot,fantasy,landscape, waterfall, tree, knight, helmet, cloudy sky, outdoors, male focus, snowing,mountain, floating island, cloud, scenery, sky, day, 1girl, no humans, fire,nature, sword, castle, armor, cape, forest, lantern, grass, water, light rays, flower, sunlight, bridge, fantasy, house,villages, rivers, buildings等等一些跟魔兽风格有关的提示词；小伙伴们，试试不同的提示词出图（推荐2-8个提示词组合），不用把上述提示词全部复制到Stable Diffusion里面，去尝试一些有趣的组合玩法，以生成不同的类似魔兽世界风格图片，3、推荐负面提示词：NSFW,(ng_deepnegative_v1_75t:1.5),(easynegative:1.5),(worst quality:2),(low quality:2),(normal quality:2),lowres,low resolution,bad anatomy,bad hands,extra fingers,missing finger,missing fingers,bad feet,extra feet,((monochrome)),((grayscale)),watermark,number,4、鉴于此模型V1.0版本原素材有限，如果只用同一个模型生成图片，不免生成的图片花样并不太多，你可尝试切换不同底模变换一下风格（比如动漫风模型：<a target="_blank" rel="ugc" href="https://civarchive.com/models/4468/counterfeit-v30">Counterfeit-V3.0</a>），体验一下在加载不同底模情况下得到的更多魔兽风格场景和人物，这也是一件趣事；后期在有时间的情况下，我会不断迭代补充训练，完善此模型，5、请注意：此模型并不能生成原版的魔兽图片，版权问题大家都懂，无需多说；只能生成魔兽类似风格图片，若需要商用，请谨慎，6、关于此模型的使用问题，欢迎大家贴图反馈；若有关于Stable Diffusion的使用和模型训练（炼丹）的任何问题，可一起交流，大家一起玩AI绘画，共同进步；暗号VX：fei_geer。///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////<h3 id="heading-1026">I.Overview of the model:</h3>To this day, we still miss the continent of Azeroth; vaguely remember that 10 years ago, I chose the Necromancer and started a journey to World of Warcraft; once upon a time, in order to complete a series of achievements, I forgot to eat and sleep, in order to form a team Set the alarm clock regularly in the next copy, in order to catch a baby mount, squatting all night... The original intention of training this model is to commemorate the passing green years. All training materials come from Internet pictures, game screenshots, etc.; you can try to use This model generates a picture as a wallpaper, look through it occasionally, and commemorate your World of Warcraft.<h3 id="heading-1027">II.The model use instructions:</h3>1, Recommended lora weight: 0.7-0.9, add trigger word: wow2wow.2, Recommended to load the base large model: <a target="_blank" rel="ugc" href="https://civarchive.com/models/7371?modelVersionId=43978">ReV Animated_v1.2.2</a>.3, The number of iteration steps: 25-40, CFG scale: 10-15, Clip skip: 2.4, Recommended sampler: DPM++ SDE Karras.5, VAE: <a target="_blank" rel="ugc" href="https://huggingface.co/mmmoof1/vae-ft-mse-840000-ema-pruned/blob/main/vae-ft-mse-840000-ema-pruned.safetensors">vae-ft-mse-840000-ema-pruned</a>.6, For the resolution settings, the recommended horizontal version: 768x512 (or 960x768), vertical cell phone size: 512x768 (or 768x960), after getting the above picture, you can subsequently use HD repair (redrawing amplitude 0.3-0.4, zoom algorithm R-ESRGAN 4x +) zoom 2 times to increase the rich picture details; of course, you can also choose Instead of HD restoration, you can use Tiled Diffusion + Tiled VAE to enlarge the image, which will give you a better quality image.7, The generalization of this model is not bad, and it can produce pictures under different base models; for example, switch to a real-style or animation-style large model to generate some interesting pictures, friends can try.<h3 id="heading-1028">III. Other notes:</h3>1, Warcraft lora standard short call words:----------------------------------------------------------------------------------------------wow2wow,&lt;lora:wow2wow:0.9&gt;.darkness,star \(sky\), weapon,glowing.masterpiece, best quality, most detailed, best aesthetics, highest contrast, highest resolution, best details, 8K wallpapers,----------------------------------------------------------------------------------------------Load lora directly, add the trigger word:wow2wow, and then with some prompt words you can get some pictures of scenes and characters in the World of Warcraft style type.2, Some other recommended prompt words:----------------------------------------------------------------------------------------------darkness,weapon,star\(sky\),glowing,android,robot,fantasy,landscape,waterfall,tree,knight,helmet,cloudy sky,outdoors,male focus. snowing,mountain, floating island, cloud, scenery, sky, day, 1girl, no humans, fire,nature, sword, castle, armor, cape, forest, lantern, grass, water, light rays, flower. sunlight, bridge, fantasy, house,villages, rivers, buildings, etc.Some prompt words related to Warcraft style,you can try them all;----------------------------------------------------------------------------------------------Guys, try different cue words out of the map (Recommended combinations of 2-8 prompt words), without copying all the above cue words into Stable Diffusion, to try some interesting combinations of play to generate different pictures similar to the World of Warcraft style.3, Recommended negative cue words:-----------------------------------------------------------------------------------------------NSFW,(ng_deepnegative_v1_75t:1.5),(easynegative:1.5),(worst quality:2),(low quality:2),(normal quality:2),lowres,low resolution,bad anatomy,bad hands,extra fingers,missing finger,missing fingers,bad feet,extra feet,((monochrome)),((grayscale)),watermark,number------------------------------------------------------------------------------------------------4, In view of the limited original materials of this model V1.0, if you only use the same model to generate pictures, the generated pictures will not have too many patterns. You can try to switch different base models to change the style (for example, anime style model: <a target="_blank" rel="ugc" href="https://civarchive.com/models/4468/counterfeit-v30">Counterfeit-V3. 0</a>), it is also interesting to experience more Warcraft-style scenes and characters obtained by loading different bottom models; later, when I have time, I will continue to iterate and supplement training to improve this model,5, Please note: this model does not generate the original Warcraft picture, copyright issues we all understand, needless to say, only similar style images of Warcraft can be generated, if you need commercial use, please be careful.6, Regarding the use of this model, everyone is welcome to map feedback;If there are any questions about the use of Stable Diffusion and model training (alchemy), you can communicate together, we play AI painting together and progress together.

魔兽风格之旅（WoW_Style_Tour）

1、这是v1.0版本，后续有时间会继续优化迭代，2、模型使用过程中，有任何问题欢迎贴图反馈，3、这个模型的泛化性还行，可以在不同的底模下出图；比如切换到真实风格或动漫风格大模型下生成一些有趣的图片，4、想要更好的生成图片，请查看详细说明。-------------------------------------------------------1, This is the v1.0 version, and we will continue to optimize iteratively when we have time.2, During the use of the model, if you have any questions, please feel free to give feedback on the texture.3, The generalization of this model is not bad, and it can produce pictures under different base models; for example, switch to a real-style or animation-style large model to generate some interesting pictures.4, If you want to generate a better picture, please check the detailed description.

Files with SHA256: 6e36b0ebc592ef3be9e33f795971e103210f6882e3fc4dc90dc865f999cc72b6