前言：这是我出的最全的一次概念： Flux &amp; SDXL &amp; SD1.5，源于我喜欢这个概念，但事后想想；虽然过程会耗费我太多的时间，但做为拥有强迫症的我认为这是一次好的开端，这代表着后续所有概念都将覆盖市场主流版本，每一位关注者都能找到适合自己的版本使用：）Preface: This is the most complete concept I have ever released: Flux &amp; SDXL &amp; SD1.5. It was because I liked this concept, but after thinking about it; although the process would take up too much of my time, as someone with obsessive-compulsive disorder, I think this is a good start, which means that all subsequent concepts will cover the mainstream versions in the market, and every follower can find a version that suits them:)<h3 id="sdxl-1.0-92ea57ssc">SDXL 1.0版本说明：</h3>我所使用到的Lora权重系数是:0.85，(我训练的Lora兼容性较好，您完全可以测试其他权重，找到适合您的为止）触发词格式参照下方Flux.1D说明。我所使用的采样器为官方推荐的DPM++ 2M SDE-Karras，采样步数:30提示词相关系数4.5即可，您可以出横图、竖图、方图、任意格式的图均可，Lora的接纳性非常好，您还是可以通过您可通过核心触发词+自然语言描绘角色所处场景+视觉角度（斜角拍摄、从上方拍摄、从正面拍摄等等...）+人物服饰/赤裸+嘴上是否保留脚步/口球。<h3 id="sdxl-1.0-version-description:-4kyedew8i">SDXL 1.0 version description:</h3>The Lora weight coefficient I used is: 0.85, (the Lora I trained is more compatible, you can test other weights until you find the one that suits you)Please refer to the Flux.1D instructions below for the trigger word format.The sampler I used is the officially recommended DPM++ 2M SDE-Karras, sampling steps: 30The prompt word correlation coefficient is 4.5, you can produce horizontal, vertical, square, or any format of pictures, Lora is very receptive, you can still use the core trigger word + natural language to describe the scene where the character is in + visual angle (angle shooting, shooting from above, shooting from the front, etc.) + character clothing/naked + whether the footsteps/ball mouth are retained on the mouth.基于VAE模型训练的下载链接(Download link based on VAE model training)：<a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/sdxl-vae/resolve/main/sdxl_vae.safetensors">sdxl_vae.safetensors</a><hr /><h3 id="flux.1d-wnpf6519t">Flux.1D版本说明：</h3>Lora权重系数可参考0.8-1（您也可以测试其他权重，找到适合您的为止）您可通过核心触发词+自然语言描绘角色所处场景+视觉角度（斜角拍摄、从上方拍摄、从正面拍摄等等...）+人物服饰/赤裸+嘴上是否保留脚步/口球.触发语言格式讲解：<pre><code>触发语言格式讲解：（女子俯卧在地，双手背在身后，双腿弯曲抬起，身上被绳索捆绑，斜角拍摄，女子俯卧在地，身上被白色绳索捆绑，双手双脚绑在一起）+（嘴上是否绑胶带/脸上表情）+（她身穿的衣服和裤子/赤裸）+（鞋子的颜色、类别、赤脚）+（所在场景）</code></pre>触发语言格式示范：<pre><code>女子俯卧在地，双手背在身后，双腿弯曲抬起，身上被绳索捆绑，斜角拍摄，女子俯卧在地，身上被白色绳索捆绑，双手双脚绑在一起，嘴上贴着银色胶带。她身穿蓝色上衣和灰色裙子，脚穿黑色高跟鞋，背景是带有窗户的深色室内空间</code></pre>我所使用的采样器为官方推荐的euler，调度器：simple，步数：50-80（当采样步数&gt;=80，生图期间平均会消耗40GB运行内存，您可以根据自己的电脑配置来调节步数）我主页发布的Flux-Lora出图非常稳定，关于network_dim和network_alpha的设定值逻辑和SD1.5及SDXL有所不同，并不是文件越大效果越好，各位关注者可以放心下载使用。<h3 id="flux.1d-version-description:-qajmvalft">Flux.1D version description:</h3>Lora coefficient weights can refer to 0.8-1 (you can also test other weights until you find the one that suits you)You can use core trigger words + natural language to describe the scene where the character is in + visual angle (angle shot, shot from above, shot from the front, etc.) + character clothing/naked + whether there are footsteps/ball on the mouth.Trigger language format explanation:<pre><code>Trigger language format explanation: (Woman lying prone on the ground, hands behind her back, legs bent and raised, tied with ropes, angled shot, woman lying prone on the ground, tied with white ropes, hands and feet tied together) + (whether there is tape on the mouth/facial expression) + (the clothes and pants she wears/naked) + (shoe color, category, barefoot) + (scene)</code></pre>Trigger language format demonstration:<pre><code>Woman lying prone on the ground, hands behind her back, legs bent and raised, tied with ropes, angled shot, woman lying prone on the ground, tied with white ropes, hands and feet tied together, silver tape on her mouth. She is wearing a blue top and a gray skirt, black high heels, and the background is a dark interior space with windows</code></pre>The sampler I use is the officially recommended euler, scheduler: simple, step number: 50-80(When the sampling step number is &gt;= 80, an average of 40GB of running memory will be consumed during the image generation period. You can adjust the step number according to your computer configuration)The Flux-Lora image posted on my homepage is very stable. The setting logic of network_dim and network_alpha is different from SD1.5 and SDXL. It is not the case that the larger the file, the better the effect. Followers can download and use it with confidence.如果你喜欢我的Lora，在决定是否使用“buzz”来鼓励我之前，你可以仔细查看我的主页作品。祝君出图愉快！If you like my Lora, you can take a closer look at my homepage works before deciding whether to use "buzz" to encourage me. I wish you happy drawing!<hr /><h3 id="sd1.5sd1.5-version-description-t5qtyrkm1">SD1.5版本说明（SD1.5 version description）：</h3>关于使用绳索实现女性直臂和肘部结合的艺术，请使用自然语言提示进行描述，以获得更好的效果，可以从正面和侧面呈现。Regarding the art of using ropes to achieve the combination of straight arms and elbows in women, please use natural language prompts to describe it for better results, which can be presented from both the front and side。这款Lora不太依赖于定制的基础模型，适用于绝大多数现实世界的模型。每个人都可以自由创造。This Lora does not rely too much on customized base models and is suitable for the vast majority of real-world models. Everyone is free to create。我的创作风格很少涉及Civitai平台上的重复作品，质量很高。如果你喜欢我的作品，请给我留下关注，我会在未来创造更多的色情艺术风格。My creative style rarely involves repetitive works on the Civitai platform, and the quality is very high. If you like my works, please leave me a follow, and I will create more erotic art styles in the future。如果你喜欢我的lora，可以找我讨论；我的电报ID：@Jonathan_Madison。If you like my Lora, you can discuss with me;My Telegram ID:@ Jonathan_Madison。

直臂并肘-bondage art-  SDXL 1.0 & Flux.1D & SD1.5

Files with SHA256: ec2cdafd7615465795d7f94952ba5c438d1d804a06b0640948169cd940a09d61