rebuilding site 2024年10月 9日星期三 14时42分56秒 CST

OldJii · Oct 9, 2024 · 8b9f522 · 8b9f522
1 parent aec0d78
commit 8b9f522
Show file tree

Hide file tree

Showing 38 changed files with 11,967 additions and 685 deletions.
diff --git a/categories/index.xml b/categories/index.xml
@@ -1,23 +1,20 @@
 <?xml version="1.0" encoding="utf-8" standalone="yes"?>
-<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
+<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
   <channel>
     <title>Categories on Folay‘s Blog</title>
     <link>https://folay.top/categories/</link>
-    <description>Recent content in Categories on Folay‘s Blog</description>
+    <description>Recent content in Categories on Folay‘s Blog feedId:66687784300023808+userId:66421915972268032</description>
     <generator>Hugo -- gohugo.io</generator>
     <language>zh-CN</language>
-    <lastBuildDate>Mon, 18 Mar 2024 15:49:06 +0800</lastBuildDate>
-
-	<atom:link href="https://folay.top/categories/index.xml" rel="self" type="application/rss+xml" />
-
-
+    <lastBuildDate>Mon, 18 Mar 2024 15:49:06 +0800</lastBuildDate><atom:link href="https://folay.top/categories/index.xml" rel="self" type="application/rss+xml" />
     <item>
       <title>生活</title>
       <link>https://folay.top/categories/%E7%94%9F%E6%B4%BB/</link>
       <pubDate>Mon, 18 Mar 2024 15:49:06 +0800</pubDate>
 
       <guid>https://folay.top/categories/%E7%94%9F%E6%B4%BB/</guid>
       <description></description>
+      <content:encoded><![CDATA[]]></content:encoded>
     </item>
 
     <item>
@@ -27,7 +24,8 @@
 
       <guid>https://folay.top/categories/%E5%B7%A5%E4%BD%9C/</guid>
       <description></description>
+      <content:encoded><![CDATA[]]></content:encoded>
     </item>
 
   </channel>
-</rss>
+</rss>
diff --git a/categories/工作/index.xml b/categories/工作/index.xml
diff --git a/categories/生活/index.xml b/categories/生活/index.xml
diff --git a/index.xml b/index.xml
diff --git a/post/index.xml b/post/index.xml
diff --git a/tags/AICG/index.xml b/tags/AICG/index.xml
@@ -1,16 +1,12 @@
 <?xml version="1.0" encoding="utf-8" standalone="yes"?>
-<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
+<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
   <channel>
     <title>AICG on Folay‘s Blog</title>
     <link>https://folay.top/tags/AICG/</link>
-    <description>Recent content in AICG on Folay‘s Blog</description>
+    <description>Recent content in AICG on Folay‘s Blog feedId:66687784300023808+userId:66421915972268032</description>
     <generator>Hugo -- gohugo.io</generator>
     <language>zh-CN</language>
-    <lastBuildDate>Tue, 26 Sep 2023 15:49:06 +0800</lastBuildDate>
-
-	<atom:link href="https://folay.top/tags/AICG/index.xml" rel="self" type="application/rss+xml" />
-
-
+    <lastBuildDate>Tue, 26 Sep 2023 15:49:06 +0800</lastBuildDate><atom:link href="https://folay.top/tags/AICG/index.xml" rel="self" type="application/rss+xml" />
     <item>
       <title>Stable Diffusion 的光影艺术</title>
       <link>https://folay.top/post/sd_controlnet/</link>
@@ -21,7 +17,43 @@
 &lt;div align=&#34;center&#34;&gt;
 &lt;img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/picgo/202310192146502.png width=100%/&gt;
 &lt;/div&gt;</description>
+      <content:encoded><![CDATA[<p>最近在网上经常有看到类似上面这种图片，图片里的建筑物或者是光影形成了一些文字，给图片附加了更多的信息。评论有很多人很好奇，所以记录下制作方式。</p>
+<div align="center">
+<img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/picgo/202310192146502.png width=100%/>
+</div>
+<p>其实制作的方法很简单，只是在基本 text2img 的基础上使用 ControlNet 进行垫图并附加一些光影性强的 Model 即可。</p>
+<h2 id="checkpoint-配置">Checkpoint 配置</h2>
+<ul>
+<li>Checkpoint Model：推荐使用写实类的模型，比如 majicMIX realistic、dalcefoPainting 3rd。</li>
+<li>Sampler method：推荐使用 DMP++ 2S a Karras、DMP++ 2M Karras。</li>
+<li>Sampling Steps：推荐 25 左右，不要超过 30，因为要实现光影文字效果，Step 数太多会渲染掉文-字效果。</li>
+</ul>
+<div align="center">
+<img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/picgo/202310192143384.png width=100%/>
+</div>
+<h2 id="底图制作">底图制作</h2>
+<p>需要用到一张黑底白字的图片，用什么做随意，用你熟悉工具即可，尺寸需要跟你上面配置的 Size 信息保持一致，以保证文字的位置是准确的，比如下图。</p>
+<div align="center">
+<img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/picgo/202310192150418.png width=100%/>
+</div>
+<p>另外，如果你能让文字的边缘实现模糊效果就更好了，生成的图片光影文字效果会更贴合。</p>
+<h2 id="controlnet-配置">ControlNet 配置</h2>
+<ul>
+<li>ControlNet Model：对出图效果的影响比较关键，推荐使用 control_v1p_sd15_brightness、lightingBasedPicture_v10，都是比较优秀的光影 Model。</li>
+<li>ControlNet Type：全部</li>
+<li>预处理器：推荐 none，也可以使用 tile_colorfix，但是要搭配 control_v11f1e_sd15_tile 模型。</li>
+<li>Control Weight：0.4 ～ 0.55</li>
+<li>Starting Control Step：0 ~ 0.1</li>
+<li>Ending Control Step：0.45 ～ 0.6</li>
+</ul>
+<div align="center">
+<img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/picgo/202310192144711.png width=100%/>
+</div>
+<p>最后就抽卡吧，慢慢抽，另外附上一张网友测试的 Control Step 与出图效果的对比图。</p>
+<div align="center">
+<img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/picgo/202310192154764.webp width=100%/>
+</div>]]></content:encoded>
     </item>
 
   </channel>
-</rss>
+</rss>
diff --git a/tags/Ai-绘画/index.xml b/tags/Ai-绘画/index.xml
@@ -1,16 +1,12 @@
 <?xml version="1.0" encoding="utf-8" standalone="yes"?>
-<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
+<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
   <channel>
     <title>Ai 绘画 on Folay‘s Blog</title>
     <link>https://folay.top/tags/Ai-%E7%BB%98%E7%94%BB/</link>
-    <description>Recent content in Ai 绘画 on Folay‘s Blog</description>
+    <description>Recent content in Ai 绘画 on Folay‘s Blog feedId:66687784300023808+userId:66421915972268032</description>
     <generator>Hugo -- gohugo.io</generator>
     <language>zh-CN</language>
-    <lastBuildDate>Wed, 22 Feb 2023 15:49:06 +0800</lastBuildDate>
-
-	<atom:link href="https://folay.top/tags/Ai-%E7%BB%98%E7%94%BB/index.xml" rel="self" type="application/rss+xml" />
-
-
+    <lastBuildDate>Wed, 22 Feb 2023 15:49:06 +0800</lastBuildDate><atom:link href="https://folay.top/tags/Ai-%E7%BB%98%E7%94%BB/index.xml" rel="self" type="application/rss+xml" />
     <item>
       <title>在 Colab 搭建 Stable Diffusion</title>
       <link>https://folay.top/post/stable_diffusion/</link>
@@ -21,7 +17,55 @@
 &lt;img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/sdsm.png width=600 /&gt;
 &lt;/div&gt;
 &lt;p&gt;Form &lt;a href=&#34;https://weirdwonderfulai.art/resources/disco-diffusion-70-plus-artist-studies&#34;&gt;Weird Wonderful Ai Art&lt;/a&gt;&lt;/p&gt;</description>
+      <content:encoded><![CDATA[<div align="center">
+<img src=https://imgoldjii.oss-cn-beijing.aliyuncs.com/sdsm.png width=600 />
+</div>
+<p>Form <a href="https://weirdwonderfulai.art/resources/disco-diffusion-70-plus-artist-studies">Weird Wonderful Ai Art</a></p>
+<blockquote>
+<p>2023 年 9 月 23 日更新：Colab 官方禁用 Stable Diffusion WebUI，无法解决断网问题，本文方法作废。详情见 <a href="https://decrypt.co/197428/google-colab-stable-diffusion-web-ui-ban">https://decrypt.co/197428/google-colab-stable-diffusion-web-ui-ban</a></p>
+</blockquote>
+<h2 id="-ai-绘画">🎨 Ai 绘画</h2>
+<p>Stable Diffusion 是 stability.ai 于 2022 年发布的深度学习文生图模型，可以在大多数配备有适度 GPU 的电脑硬件上运行。</p>
+<p>Colaboratory 简称 Colab，是 Google Research 团队发布的一种托管式 Jupyter 笔记本服务，可以免费使用 GPU / TPU 计算资源。</p>
+<p>这是一份保姆级的在 Colab 搭建 Stable Diffusion 服务的教程，按照以下步骤操作，即可浅尝 Ai 绘画的魅力。</p>
+<h2 id="分配资源">分配资源</h2>
+<p>进入 <a href="https://colab.research.google.com/github/altryne/sd-webui-colab/blob/main/Stable_Diffusion_WebUi_Altryne.ipynb">Stable_Diffusion_WebUi_Altryne</a> Colab 工程。</p>
+<p>点击右上角的<strong>连接</strong>获取云主机资源并连接运行时。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222142539642.png" alt="1-image-20230222142539642"></p>
+<p>运行 1.0 单元格中的命令可以获主机信息。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222142936172.png" alt="1-image-20230222142936172"></p>
+<p>一般分配到的都是 Tesla T4 机型，如果人品欧分配到 V100，那可以好好的吹一波了。</p>
+<h2 id="配置模型">配置模型</h2>
+<p>首先注册 <a href="https://huggingface.co/">Hugging Face</a> 账号，并生成 WRITE 类型的 <a href="https://huggingface.co/settings/tokens">Access Tokens</a>。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222144043327.png" alt="1-image-20230222144043327"></p>
+<p>返回到 Colab 界面，展开 <em>1.4 Connect to Google Drive</em> 单元格，勾选 download_if_missing 并输入 Hugging Face 的 Accsee Token。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222144232989.png" alt="1-image-20230222144232989"></p>
+<h2 id="设置密码">设置密码</h2>
+<p>在 <em>2.1 Optional - Set webUI settings and configs before running</em> 中配置密码。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222144551905.png" alt="1-image-20230222144551905"></p>
+<h2 id="运行服务">运行服务</h2>
+<p>将所有单元块收起，依次运行所有单元块部署服务即可。</p>
+<p>运行耗时</p>
+<ul>
+<li>Setup stage：5min</li>
+<li>Run the Stable Diffusion webui：1s</li>
+<li>Launch WebUI for stable diffusion：1min</li>
+</ul>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222144857944.png" alt="1-image-20230222144857944"></p>
+<p>运行过程中会弹窗请求登录 Google Drive，正常登录并允许即可。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222145720360.png" alt="1-image-20230222145720360"></p>
+<p>运行完成点击 Public URL 进入 Stable Diffusion WebUI 界面。账号为 wubei，密码为上面自己配置的。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222150541478.png" alt="1-image-20230222150541478"></p>
+<p>然后就可以尽情的享用了。</p>
+<p>输入 prompt，即可生成对应的图像，比如输入 “ an astronaut in the water ”。</p>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222151132661.png" alt="1-image-20230222151132661"></p>
+<p>当然，你可以输入更多的 prompt 信息，来生成更加具体准确的图像。</p>
+<blockquote>
+<p>(((masterpiece))),((best quality)), flat chest,((loli)),((one girl)),very long light white hair, beautiful detailed red eyes,aqua eyes,white robe, cat ears,(flower hairpin),sunlight, light smile,blue necklace,see-through,</p>
+<p>杰作，最佳品质，贫乳，萝莉，1个女孩，很长的头发，淡白色头发，红色眼睛，浅绿色眼睛，白色长裙，猫耳，发夹，阳光下，淡淡的微笑，蓝色项链，透明</p>
+</blockquote>
+<p><img src="https://imgoldjii.oss-cn-beijing.aliyuncs.com/1-image-20230222154010936.png" alt="1-image-20230222154010936"></p>]]></content:encoded>
     </item>
 
   </channel>
-</rss>
+</rss>