<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>PromptZone - AI Prompts, Guides and Tools for Builders: celine</title>
    <description>The latest articles on PromptZone - AI Prompts, Guides and Tools for Builders by celine (@celine).</description>
    <link>https://www.promptzone.com/celine</link>
    <image>
      <url>https://promptzone-community.s3.amazonaws.com/uploads/user/profile_image/7543/13fabcda-700f-4ff3-96d1-21982aac6e1b.jpg</url>
      <title>PromptZone - AI Prompts, Guides and Tools for Builders: celine</title>
      <link>https://www.promptzone.com/celine</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://www.promptzone.com/feed/celine"/>
    <language>en</language>
    <item>
      <title>Install ComfyUI (2026): Windows, Mac, Linux and Desktop App</title>
      <dc:creator>celine</dc:creator>
      <pubDate>Sat, 22 Feb 2025 09:17:10 +0000</pubDate>
      <link>https://www.promptzone.com/celine/comfyui-installation-guide-a-comprehensive-tutorial-56h</link>
      <guid>https://www.promptzone.com/celine/comfyui-installation-guide-a-comprehensive-tutorial-56h</guid>
      <description>&lt;p&gt;&lt;strong&gt;Windows:&lt;/strong&gt; Install Comfy Desktop for a guided setup, or extract the portable build matched to your GPU.&lt;br&gt;&lt;br&gt;
&lt;strong&gt;Mac:&lt;/strong&gt; Install Comfy Desktop on Apple Silicon; use the manual route when you need control over Python and PyTorch.&lt;br&gt;&lt;br&gt;
&lt;strong&gt;Linux:&lt;/strong&gt; Use the official Desktop AppImage or Debian package, or install manually for a custom GPU environment.&lt;/p&gt;
&lt;h2 id="overview"&gt;
  
  
  Overview
&lt;/h2&gt;

&lt;p&gt;ComfyUI connects models and generation steps through a node graph. This guide covers five installation routes and your first local image.&lt;/p&gt;

&lt;p&gt;The &lt;a href="https://docs.comfy.org/installation/desktop/overview" rel="nofollow ugc noopener noreferrer"&gt;Desktop overview&lt;/a&gt; lists Windows, Apple Silicon macOS, and Linux packages. The core README's quick-start summary lists only Windows and macOS; use the platform-specific Desktop instructions.&lt;/p&gt;

&lt;p&gt;For manual installations, the &lt;a href="https://docs.comfy.org/installation/system_requirements" rel="nofollow ugc noopener noreferrer"&gt;system requirements&lt;/a&gt; recommend Python &lt;strong&gt;3.13&lt;/strong&gt;, with &lt;strong&gt;3.12&lt;/strong&gt; as a fallback for custom-node dependency problems. Choose your installation method before adding extensions.&lt;/p&gt;
&lt;h2 id="which-install-method"&gt;
  
  
  Which install method
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Method&lt;/th&gt;
&lt;th&gt;Platforms&lt;/th&gt;
&lt;th&gt;Choose it when&lt;/th&gt;
&lt;th&gt;Main consideration&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Comfy Desktop&lt;/td&gt;
&lt;td&gt;Windows, Apple Silicon Mac, Linux&lt;/td&gt;
&lt;td&gt;You want guided installation and environment management&lt;/td&gt;
&lt;td&gt;Each instance has its own configuration&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Portable package&lt;/td&gt;
&lt;td&gt;Windows&lt;/td&gt;
&lt;td&gt;You want an extracted folder with bundled Python&lt;/td&gt;
&lt;td&gt;Match the package to your GPU&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Manual installation&lt;/td&gt;
&lt;td&gt;Windows, macOS, Linux&lt;/td&gt;
&lt;td&gt;You need direct control over dependencies&lt;/td&gt;
&lt;td&gt;Maintain the Python environment yourself&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;comfy-cli&lt;/td&gt;
&lt;td&gt;Windows, macOS, Linux&lt;/td&gt;
&lt;td&gt;You prefer terminal-based installation and launching&lt;/td&gt;
&lt;td&gt;GPU prerequisites still matter&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Stability Matrix&lt;/td&gt;
&lt;td&gt;Windows, macOS, Linux&lt;/td&gt;
&lt;td&gt;You manage several generation applications&lt;/td&gt;
&lt;td&gt;Follow its package and model management conventions&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;
&lt;h2 id="methods"&gt;
  
  
  Installation Methods
&lt;/h2&gt;

&lt;p&gt;For a first installation, I recommend Desktop on supported hardware. Choose portable or manual installation when you want direct access to the underlying environment.&lt;/p&gt;
&lt;h3 id="1-comfy-desktop-app"&gt;
  
  
  1. Comfy Desktop app
&lt;/h3&gt;

&lt;p&gt;Desktop creates ComfyUI instances and manages their Python dependencies. Its &lt;a href="https://docs.comfy.org/installation/desktop/usage/overview" rel="nofollow ugc noopener noreferrer"&gt;usage guide&lt;/a&gt; describes independent instances and shared model storage.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Windows:&lt;/strong&gt; Get the installer through the &lt;a href="https://docs.comfy.org/installation/desktop/windows" rel="nofollow ugc noopener noreferrer"&gt;official Windows guide&lt;/a&gt;, run the &lt;code&gt;.exe&lt;/code&gt;, and launch Comfy Desktop. The page lists Windows 10 or later, with a dedicated NVIDIA or AMD GPU recommended.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Mac:&lt;/strong&gt; Follow the &lt;a href="https://docs.comfy.org/installation/desktop/macos" rel="nofollow ugc noopener noreferrer"&gt;macOS guide&lt;/a&gt;: open the &lt;code&gt;.dmg&lt;/code&gt;, drag the app into Applications, and launch it. Desktop requires Apple Silicon and macOS 13 or later; Intel Macs are outside this Desktop route.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Linux:&lt;/strong&gt; The &lt;a href="https://docs.comfy.org/installation/desktop/linux" rel="nofollow ugc noopener noreferrer"&gt;Linux guide&lt;/a&gt; provides &lt;code&gt;.deb&lt;/code&gt; and AppImage packages for x64, recommending Ubuntu 22.04 or later. Download into a folder containing only the chosen installer.&lt;/p&gt;

&lt;p&gt;For Debian/Ubuntu, run from that folder:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;sudo &lt;/span&gt;apt &lt;span class="nb"&gt;install&lt;/span&gt; ./&lt;span class="k"&gt;*&lt;/span&gt;.deb
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Alternatively, run the downloaded AppImage:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;chmod&lt;/span&gt; +x ./&lt;span class="k"&gt;*&lt;/span&gt;.AppImage
./&lt;span class="k"&gt;*&lt;/span&gt;.AppImage
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Create your first installation from Desktop's welcome screen. Let dependency setup finish, then launch ComfyUI through the instance controls.&lt;/p&gt;

&lt;p&gt;Desktop updates and ComfyUI-instance updates are separate: the Desktop Update control updates the launcher, while the instance's Manage panel handles its ComfyUI update.&lt;/p&gt;

&lt;h3 id="2-installation-via-the-precompiled-package"&gt;
  
  
  2. Installation via the Pre-Compiled Package
&lt;/h3&gt;

&lt;p&gt;The &lt;a href="https://docs.comfy.org/installation/comfyui_portable_windows" rel="nofollow ugc noopener noreferrer"&gt;Windows portable guide&lt;/a&gt; documents separate NVIDIA, AMD, and Intel packages. Extract the archive completely, then run the matching launcher from its root folder.&lt;/p&gt;

&lt;p&gt;For NVIDIA, in PowerShell:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;\run_nvidia_gpu.bat&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;AMD and Intel packages provide &lt;code&gt;run_amd_gpu.bat&lt;/code&gt; and &lt;code&gt;run_intel_gpu.bat&lt;/code&gt;; CPU execution uses &lt;code&gt;run_cpu.bat&lt;/code&gt;. Keep the terminal open while using the browser interface.&lt;/p&gt;

&lt;p&gt;The documented NVIDIA builds distinguish CUDA 13.0/Python 3.13 for RTX hardware from CUDA 12.6/Python 3.12 for NVIDIA 10-series and older hardware. Read the package notes before downloading.&lt;/p&gt;

&lt;p&gt;When startup prints its local address, open &lt;a href="http://localhost:8188" rel="nofollow ugc noopener noreferrer"&gt;http://localhost:8188&lt;/a&gt; if the browser does not open automatically. The bundle's Python folder is named &lt;strong&gt;&lt;code&gt;python_embeded&lt;/code&gt;&lt;/strong&gt;, including that spelling.&lt;/p&gt;

&lt;p&gt;The folder is portable, but another machine still needs compatible hardware and drivers.&lt;/p&gt;

&lt;h3 id="3-manual-installation"&gt;
  
  
  3. Manual Installation
&lt;/h3&gt;

&lt;p&gt;Install Git and Miniconda first. On Windows, use a terminal where Conda is initialized. The &lt;a href="https://docs.comfy.org/installation/manual_install" rel="nofollow ugc noopener noreferrer"&gt;manual guide&lt;/a&gt; also calls for the Microsoft Visual C++ Redistributable on Windows.&lt;/p&gt;

&lt;p&gt;Create an isolated environment using &lt;a href="https://docs.conda.io/projects/conda/en/stable/user-guide/tasks/manage-environments.html" rel="nofollow ugc noopener noreferrer"&gt;Conda's documented environment commands&lt;/a&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;conda create &lt;span class="nt"&gt;-n&lt;/span&gt; comfyenv &lt;span class="nv"&gt;python&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;3.13
conda activate comfyenv
git clone https://github.com/Comfy-Org/ComfyUI.git
&lt;span class="nb"&gt;cd &lt;/span&gt;ComfyUI
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Keep this environment active and install &lt;strong&gt;one&lt;/strong&gt; PyTorch build matching your hardware. The original &lt;code&gt;comfyanonymous/ComfyUI&lt;/code&gt; address redirects to Comfy-Org.&lt;/p&gt;

&lt;h4 id="windows-or-linux-with-nvidia"&gt;
  
  
  Windows or Linux with NVIDIA
&lt;/h4&gt;

&lt;p&gt;The &lt;a href="https://github.com/comfyanonymous/ComfyUI" rel="nofollow ugc noopener noreferrer"&gt;ComfyUI README&lt;/a&gt; currently supplies this stable CUDA command:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-m&lt;/span&gt; pip &lt;span class="nb"&gt;install &lt;/span&gt;torch torchvision torchaudio &lt;span class="nt"&gt;--extra-index-url&lt;/span&gt; https://download.pytorch.org/whl/cu130
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Use a compatible NVIDIA driver. The README requires a cu130-or-newer PyTorch build for NVIDIA 20-series and newer hardware; its older-GPU portable package is a separate route.&lt;/p&gt;

&lt;h4 id="linux-with-amd"&gt;
  
  
  Linux with AMD
&lt;/h4&gt;

&lt;p&gt;For a supported AMD GPU and ROCm environment, the manual guide supplies the ROCm 7.2 build:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-m&lt;/span&gt; pip &lt;span class="nb"&gt;install &lt;/span&gt;torch torchvision torchaudio &lt;span class="nt"&gt;--index-url&lt;/span&gt; https://download.pytorch.org/whl/rocm7.2
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Check your GPU and operating system against AMD's compatibility documentation linked from the ComfyUI README. Installing a wheel cannot make an unsupported hardware combination compatible.&lt;/p&gt;

&lt;p&gt;For &lt;strong&gt;AMD on Windows&lt;/strong&gt;, the official guide now documents ROCm packages as well. Use the AMD portable package for the simpler route, or follow its hardware-specific manual instructions.&lt;/p&gt;

&lt;h4 id="mac-with-apple-silicon"&gt;
  
  
  Mac with Apple Silicon
&lt;/h4&gt;

&lt;p&gt;Apple's &lt;a href="https://developer.apple.com/metal/pytorch/" rel="nofollow ugc noopener noreferrer"&gt;PyTorch guide&lt;/a&gt; explains MPS acceleration and lists the current PyTorch requirements, including macOS 14 or later. That PyTorch requirement is distinct from Desktop's launcher requirement.&lt;/p&gt;

&lt;p&gt;Install Apple's command-line tools if needed:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;xcode-select &lt;span class="nt"&gt;--install&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;ComfyUI's README recommends nightly PyTorch on Apple Silicon. Apple's documented nightly command, run in the active environment, is:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-m&lt;/span&gt; pip &lt;span class="nb"&gt;install&lt;/span&gt; &lt;span class="nt"&gt;--pre&lt;/span&gt; torch torchvision torchaudio &lt;span class="nt"&gt;--extra-index-url&lt;/span&gt; https://download.pytorch.org/whl/nightly/cpu
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Despite &lt;code&gt;cpu&lt;/code&gt; in the index URL, this is Apple's published macOS nightly installation route. Confirm MPS availability using the check below.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/nymlsul3mnjwv34i9fof.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/nymlsul3mnjwv34i9fof.jpg" alt="Mac Terminal Installation"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h4 id="finish-on-every-platform"&gt;
  
  
  Finish on every platform
&lt;/h4&gt;

&lt;p&gt;From the cloned ComfyUI folder, install its dependencies and start the server:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-m&lt;/span&gt; pip &lt;span class="nb"&gt;install&lt;/span&gt; &lt;span class="nt"&gt;-r&lt;/span&gt; requirements.txt
python main.py
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Open the address printed in the terminal. For later launches, activate &lt;code&gt;comfyenv&lt;/code&gt;, return to the same folder, and run &lt;code&gt;python main.py&lt;/code&gt; again. Stop the server with &lt;code&gt;Ctrl+C&lt;/code&gt;.&lt;/p&gt;

&lt;h3 id="4-install-with-comfycli"&gt;
  
  
  4. Install with comfy-cli
&lt;/h3&gt;

&lt;p&gt;The official &lt;a href="https://docs.comfy.org/comfy-cli/getting-started" rel="nofollow ugc noopener noreferrer"&gt;comfy-cli guide&lt;/a&gt; documents installation, environment inspection, and launching. In an isolated Python environment, run:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;pip &lt;span class="nb"&gt;install &lt;/span&gt;comfy-cli
comfy &lt;span class="nb"&gt;install
&lt;/span&gt;comfy which
comfy &lt;span class="nb"&gt;env
&lt;/span&gt;comfy launch
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Follow the prompts and confirm the workspace. CUDA or ROCm support must still suit your machine. Use &lt;code&gt;comfy which&lt;/code&gt; before modifying an installation when you maintain several.&lt;/p&gt;

&lt;h3 id="5-installation-via-stability-matrix"&gt;
  
  
  5. Installation via Stability Matrix
&lt;/h3&gt;

&lt;p&gt;&lt;a href="https://github.com/LykosAI/StabilityMatrix" rel="nofollow ugc noopener noreferrer"&gt;Stability Matrix&lt;/a&gt; is a third-party package manager for Windows, macOS, and Linux. It embeds Git and Python and supports ComfyUI packages and shared model management.&lt;/p&gt;

&lt;p&gt;Download its platform release, extract or install it as instructed, and launch it. Add a ComfyUI package through its package interface, let setup complete, then launch that package. This route uses the app rather than manual shell commands.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/pxbe5zld3j9ubnqc0ay2.jpeg" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/pxbe5zld3j9ubnqc0ay2.jpeg" alt="Stability Matrix Interface"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Select a compatible model for your first workflow. For broader context, see our &lt;a href="https://www.promptzone.com/deepa_kowalski/ai-image-generators-2026-vheer-visualgpt-fooocus-comfyui-midjourney-more-compared-2i44"&gt;Stable Diffusion and image-generator comparison&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="after-installing"&gt;
  
  
  After installing
&lt;/h2&gt;

&lt;p&gt;Check the environment before downloading an elaborate workflow. For manual NVIDIA or AMD ROCm installs, run:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-c&lt;/span&gt; &lt;span class="s2"&gt;"import torch; print(torch.cuda.is_available())"&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;a href="https://pytorch.org/get-started/locally/" rel="nofollow ugc noopener noreferrer"&gt;PyTorch's installation guide&lt;/a&gt; uses this check for CUDA and ROCm. Expect &lt;code&gt;True&lt;/code&gt; when the supported GPU backend is accessible.&lt;/p&gt;

&lt;p&gt;On Apple Silicon, run:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-c&lt;/span&gt; &lt;span class="s2"&gt;"import torch; print(torch.backends.mps.is_available())"&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then follow ComfyUI's &lt;a href="https://docs.comfy.org/get_started/first_generation" rel="nofollow ugc noopener noreferrer"&gt;first-generation tutorial&lt;/a&gt;: load its basic workflow, download the requested checkpoint, and select it in Load Checkpoint. Click &lt;strong&gt;Run&lt;/strong&gt; and confirm that Save Image receives a result.&lt;/p&gt;

&lt;p&gt;Refresh model lists with &lt;strong&gt;R&lt;/strong&gt; after adding files. Keep the successful workflow as a baseline for checking later changes.&lt;/p&gt;

&lt;h2 id="models"&gt;
  
  
  Managing Models and Extensions
&lt;/h2&gt;

&lt;p&gt;Paths below are relative to the ComfyUI directory. They follow the project's &lt;a href="https://github.com/Comfy-Org/ComfyUI/blob/master/folder_paths.py" rel="nofollow ugc noopener noreferrer"&gt;folder definitions&lt;/a&gt;.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Model component&lt;/th&gt;
&lt;th&gt;Folder&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Complete checkpoints&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/checkpoints/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Separate diffusion weights&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/diffusion_models/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Text encoders&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/text_encoders/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;VAE&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/vae/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;LoRA&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/loras/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;ControlNet&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/controlnet/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Upscaling models&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/upscale_models/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Use the workflow's download instructions for &lt;code&gt;.safetensors&lt;/code&gt; files. A separate diffusion model is not necessarily a complete checkpoint; &lt;a href="https://docs.comfy.org/basic-concepts/models" rel="nofollow ugc noopener noreferrer"&gt;ComfyUI's model documentation&lt;/a&gt; explains that distinction.&lt;/p&gt;

&lt;p&gt;For FLUX workflows, obtain the required encoders and VAE as well as the diffusion weights. For an SDXL starting point, follow &lt;a href="https://www.promptzone.com/jaroslav/how-to-install-and-run-sdxl-models-in-comfyui-a-complete-guide-2nk2"&gt;installing and running SDXL in ComfyUI&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;To share an Automatic1111 model library, copy &lt;code&gt;extra_model_paths.yaml.example&lt;/code&gt; to &lt;code&gt;extra_model_paths.yaml&lt;/code&gt;, edit its existing path examples, and restart. Desktop exposes external model directories through its Storage controls.&lt;/p&gt;

&lt;h2 id="manager"&gt;
  
  
  Install ComfyUI Manager First
&lt;/h2&gt;

&lt;p&gt;Before adding custom extensions, check whether Manager is already available. The &lt;a href="https://docs.comfy.org/manager/install" rel="nofollow ugc noopener noreferrer"&gt;Manager installation guide&lt;/a&gt; says Desktop includes and enables it; manual and portable installations require setup.&lt;/p&gt;

&lt;p&gt;For a manual install, stop the server and run from the ComfyUI folder:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-m&lt;/span&gt; pip &lt;span class="nb"&gt;install&lt;/span&gt; &lt;span class="nt"&gt;-r&lt;/span&gt; manager_requirements.txt
python main.py &lt;span class="nt"&gt;--enable-manager&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;For portable Windows, run from the folder containing &lt;code&gt;python_embeded&lt;/code&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;\python_embeded\python.exe&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-m&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;pip&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;install&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-r&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;ComfyUI\manager_requirements.txt&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;\python_embeded\python.exe&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-s&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;ComfyUI\main.py&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;--windows-standalone-build&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;--enable-manager&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Manager helps install and update custom nodes. A workflow using only built-in nodes does not require third-party extensions. If an imported workflow needs additional nodes, install the relevant packages and restart.&lt;/p&gt;

&lt;h2 id="troubleshooting"&gt;
  
  
  Troubleshooting
&lt;/h2&gt;

&lt;p&gt;Use the error message and active environment to choose the relevant fix below.&lt;/p&gt;

&lt;h2 id="troubleshooting"&gt;
  
  
  Fixing Common Installation Errors
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;GPU check returns False:&lt;/strong&gt; Check the active environment, PyTorch build, driver, and hardware support. This result alone does not prove that you installed a CPU-only wheel.&lt;/p&gt;

&lt;p&gt;For NVIDIA's “Torch not compiled with CUDA enabled” error, the README recommends removing Torch and reinstalling the CUDA build:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python &lt;span class="nt"&gt;-m&lt;/span&gt; pip uninstall torch
python &lt;span class="nt"&gt;-m&lt;/span&gt; pip &lt;span class="nb"&gt;install &lt;/span&gt;torch torchvision torchaudio &lt;span class="nt"&gt;--extra-index-url&lt;/span&gt; https://download.pytorch.org/whl/cu130
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Missing modules or red custom nodes:&lt;/strong&gt; Read the startup import error. Install the extension's own dependencies into the Python environment running ComfyUI. A system-wide installation will not repair the portable interpreter's environment.&lt;/p&gt;

&lt;p&gt;To isolate extension problems, use the &lt;a href="https://docs.comfy.org/troubleshooting/custom-node-issues" rel="nofollow ugc noopener noreferrer"&gt;custom-node troubleshooting guide&lt;/a&gt; and start a manual install with:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python main.py &lt;span class="nt"&gt;--disable-all-custom-nodes&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Out of memory:&lt;/strong&gt; Test a smaller workflow and resolution. The &lt;a href="https://docs.comfy.org/troubleshooting/model-issues" rel="nofollow ugc noopener noreferrer"&gt;model troubleshooting guide&lt;/a&gt; documents &lt;code&gt;--lowvram&lt;/code&gt;, then &lt;code&gt;--novram&lt;/code&gt;, as options when the model exceeds available GPU memory.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Browser never opens:&lt;/strong&gt; Read the terminal for a startup failure. If it reports an occupied port, choose another using the documented &lt;a href="https://docs.comfy.org/development/comfyui-server/startup-flags" rel="nofollow ugc noopener noreferrer"&gt;startup flags&lt;/a&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python main.py &lt;span class="nt"&gt;--port&lt;/span&gt; 8189
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Models missing or incompatible:&lt;/strong&gt; Check the folder, filename, download integrity, and model family. Refresh or restart after adding files. Tensor-shape errors can indicate a mismatched VAE or encoder.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Mac generation is unexpectedly slow:&lt;/strong&gt; Run the MPS check. If it returns &lt;code&gt;False&lt;/code&gt;, resolve the Python/PyTorch/macOS combination before comparing generation speed with GPU examples.&lt;/p&gt;

&lt;h2 id="which-method"&gt;
  
  
  Which Method Should You Choose?
&lt;/h2&gt;

&lt;p&gt;Choose Desktop for guided setup, portable for bundled Windows Python, or manual installation for dependency control. Stability Matrix suits an existing package library; comfy-cli suits terminal workflows.&lt;/p&gt;

&lt;h2 id="conclusion"&gt;
  
  
  Conclusion
&lt;/h2&gt;

&lt;p&gt;Finish with one working installation, one compatible model, and one saved image. Keep that baseline workflow available while adding the tools your next project needs.&lt;/p&gt;

&lt;h2 id="faq"&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3 id="how-much-vram-do-i-need-to-install-comfyui"&gt;
  
  
  How much VRAM do I need to install ComfyUI?
&lt;/h3&gt;

&lt;p&gt;There is no single requirement covering every model. ComfyUI supports CPU execution through &lt;code&gt;--cpu&lt;/code&gt;; GPU memory needs depend on the selected workflow. Check its model requirements before downloading large weights.&lt;/p&gt;

&lt;h3 id="does-comfyui-work-on-amd-gpus"&gt;
  
  
  Does ComfyUI work on AMD GPUs?
&lt;/h3&gt;

&lt;p&gt;Yes. Official documentation provides ROCm instructions for Linux and Windows, plus an AMD Windows portable build. Support still depends on the GPU, driver, operating system, and PyTorch combination.&lt;/p&gt;

&lt;h3 id="can-comfyui-and-automatic1111-share-models"&gt;
  
  
  Can ComfyUI and Automatic1111 share models?
&lt;/h3&gt;

&lt;p&gt;Yes. Configure &lt;code&gt;extra_model_paths.yaml&lt;/code&gt; using the supplied example and restart ComfyUI. Desktop also supports external model directories, so copying every checkpoint is unnecessary.&lt;/p&gt;

&lt;h3 id="why-do-downloaded-workflows-show-missing-nodes"&gt;
  
  
  Why do downloaded workflows show missing nodes?
&lt;/h3&gt;

&lt;p&gt;An extension may be absent or may have failed to import. Check Manager and the terminal log, install the needed package and dependencies into the correct environment, then restart.&lt;/p&gt;

&lt;h2 id="sources"&gt;
  
  
  Sources
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/comfyanonymous/ComfyUI" rel="nofollow ugc noopener noreferrer"&gt;ComfyUI repository and README&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/system_requirements" rel="nofollow ugc noopener noreferrer"&gt;System requirements&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/desktop/overview" rel="nofollow ugc noopener noreferrer"&gt;Desktop overview&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/desktop/windows" rel="nofollow ugc noopener noreferrer"&gt;Desktop Windows installation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/desktop/macos" rel="nofollow ugc noopener noreferrer"&gt;Desktop macOS installation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/desktop/linux" rel="nofollow ugc noopener noreferrer"&gt;Desktop Linux installation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/desktop/usage/overview" rel="nofollow ugc noopener noreferrer"&gt;Desktop usage&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/comfyui_portable_windows" rel="nofollow ugc noopener noreferrer"&gt;Windows portable installation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/installation/manual_install" rel="nofollow ugc noopener noreferrer"&gt;Manual installation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.conda.io/projects/conda/en/stable/user-guide/tasks/manage-environments.html" rel="nofollow ugc noopener noreferrer"&gt;Conda environment management&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://developer.apple.com/metal/pytorch/" rel="nofollow ugc noopener noreferrer"&gt;Apple PyTorch and MPS guide&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/comfy-cli/getting-started" rel="nofollow ugc noopener noreferrer"&gt;comfy-cli installation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/LykosAI/StabilityMatrix" rel="nofollow ugc noopener noreferrer"&gt;Stability Matrix&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://pytorch.org/get-started/locally/" rel="nofollow ugc noopener noreferrer"&gt;PyTorch installation and checks&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/get_started/first_generation" rel="nofollow ugc noopener noreferrer"&gt;First image generation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/Comfy-Org/ComfyUI/blob/master/folder_paths.py" rel="nofollow ugc noopener noreferrer"&gt;ComfyUI folder definitions&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/basic-concepts/models" rel="nofollow ugc noopener noreferrer"&gt;Model concepts and locations&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/manager/install" rel="nofollow ugc noopener noreferrer"&gt;Manager installation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/troubleshooting/custom-node-issues" rel="nofollow ugc noopener noreferrer"&gt;Custom-node troubleshooting&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/troubleshooting/model-issues" rel="nofollow ugc noopener noreferrer"&gt;Model troubleshooting&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/development/comfyui-server/startup-flags" rel="nofollow ugc noopener noreferrer"&gt;Server startup flags&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;em&gt;Last reviewed and updated: September 2026.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>comfyui</category>
      <category>stablediffusion</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>Claude Styles (2026): Presets, Skills and Custom Writing</title>
      <dc:creator>celine</dc:creator>
      <pubDate>Thu, 28 Nov 2024 09:13:05 +0000</pubDate>
      <link>https://www.promptzone.com/celine/claude-styles-a-game-changer-for-ai-writing-complete-guide-for-2024-2025-1fnl</link>
      <guid>https://www.promptzone.com/celine/claude-styles-a-game-changer-for-ai-writing-complete-guide-for-2024-2025-1fnl</guid>
      <description>&lt;p&gt;As of September 17, 2026, Anthropic's help center directs &lt;strong&gt;Claude styles&lt;/strong&gt; customization to &lt;strong&gt;Skills&lt;/strong&gt; and account-wide &lt;strong&gt;Instructions for Claude&lt;/strong&gt;. Its personalization guide does not document a Styles selector or current preset list. Use these documented controls to create reusable writing guidance. &lt;a href="https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features" rel="nofollow ugc noopener noreferrer"&gt;Claude personalization&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/a59uerelj9tg4w6jwuio.gif" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/a59uerelj9tg4w6jwuio.gif" alt="Claude AI Interface - Styles Feature"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Historical Styles interface; Anthropic introduced the feature on November 26, 2024.&lt;/em&gt;&lt;/p&gt;

&lt;h2 id="what-happened-to-claude-styles"&gt;
  
  
  What happened to Claude Styles?
&lt;/h2&gt;

&lt;p&gt;Anthropic introduced Styles on &lt;strong&gt;November 26, 2024&lt;/strong&gt;, for customizing tone and response structure through presets or writing samples. &lt;a href="https://www.anthropic.com/news/styles" rel="nofollow ugc noopener noreferrer"&gt;Anthropic's launch announcement&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The current personalization guide lists &lt;strong&gt;Instructions for Claude&lt;/strong&gt;, &lt;strong&gt;Project instructions&lt;/strong&gt;, and &lt;strong&gt;Skills&lt;/strong&gt;. It does not establish whether every app has removed the Styles selector or automatically migrated saved styles.&lt;/p&gt;

&lt;h2 id="which-claude-styles-presets-were-documented"&gt;
  
  
  Which Claude Styles presets were documented?
&lt;/h2&gt;

&lt;p&gt;Anthropic's &lt;strong&gt;November 26, 2024&lt;/strong&gt; announcement named three presets: Concise, Formal, and Explanatory. These are &lt;strong&gt;historical presets, not a verified current menu&lt;/strong&gt;. The current personalization guide provides no equivalent Styles preset list.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Historical preset&lt;/th&gt;
&lt;th&gt;Documented purpose&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Concise&lt;/td&gt;
&lt;td&gt;Brief, direct replies&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Formal&lt;/td&gt;
&lt;td&gt;Polished communication&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Explanatory&lt;/td&gt;
&lt;td&gt;Teaching and concept explanations&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The 2024 announcement also described custom styles created from uploaded samples and instructions. For the currently documented workflow, use the Skills procedure below. &lt;a href="https://www.anthropic.com/news/styles" rel="nofollow ugc noopener noreferrer"&gt;Historical Styles announcement&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="where-are-the-writing-controls-now"&gt;
  
  
  Where are the writing controls now?
&lt;/h2&gt;

&lt;p&gt;For general communication preferences, open your account menu, select &lt;strong&gt;Settings&lt;/strong&gt;, and find &lt;strong&gt;Instructions for Claude&lt;/strong&gt;. These instructions apply across conversations. Reserve this space for preferences you want broadly applied. &lt;a href="https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features" rel="nofollow ugc noopener noreferrer"&gt;Account instructions&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For reusable task-specific writing guidance, use &lt;strong&gt;Customize → Skills&lt;/strong&gt;. Anthropic documents toggles for enabling individual skills and a &lt;strong&gt;+ → Create skill → Upload a skill&lt;/strong&gt; route for ZIP packages containing a skill folder. &lt;a href="https://support.claude.com/en/articles/12512180-use-skills-in-claude" rel="nofollow ugc noopener noreferrer"&gt;Using Skills&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="are-custom-writing-instructions-available-on-the-free-plan"&gt;
  
  
  Are custom writing instructions available on the free plan?
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Skills are available on Free, Pro, Max, Team, and Enterprise&lt;/strong&gt;, according to the current Skills overview. They require code execution to be enabled. &lt;a href="https://support.claude.com/en/articles/12512176-what-are-skills" rel="nofollow ugc noopener noreferrer"&gt;Skills availability&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For individual plans, check &lt;strong&gt;Settings → Capabilities → Code execution and file creation&lt;/strong&gt;. Team and Enterprise organizations also have organization-level controls; an owner can restrict user-created skills. &lt;a href="https://support.claude.com/en/articles/12512180-use-skills-in-claude" rel="nofollow ugc noopener noreferrer"&gt;Skills prerequisites&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="create-a-custom-writing-style-from-samples"&gt;
  
  
  Create a custom writing style from samples
&lt;/h2&gt;

&lt;p&gt;Build a writing skill around a format you regularly use. Anthropic's Skills tutorial recommends naming the format, describing the tone, and providing an example that meets your standards. &lt;a href="https://academy.claude.com/tutorials/teach-claude-your-way-of-working-using-skills" rel="nofollow ugc noopener noreferrer"&gt;Skills tutorial&lt;/a&gt;.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Choose a sample you would be happy to publish again.&lt;/li&gt;
&lt;li&gt;Identify what to preserve: sentence length, level of detail, headings, or vocabulary.&lt;/li&gt;
&lt;li&gt;Paste the sample into a conversation and ask Claude to create a skill for that writing task.&lt;/li&gt;
&lt;li&gt;Review the proposed instructions before packaging them.&lt;/li&gt;
&lt;li&gt;Upload the skill through &lt;strong&gt;Customize → Skills&lt;/strong&gt; and enable it.&lt;/li&gt;
&lt;li&gt;Try a new assignment and compare the response with your sample.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Here is an original starter prompt:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Create a reusable writing skill from the sample below. Describe its tone, paragraph structure, vocabulary, and level of technical explanation. Preserve those patterns without copying the sample's facts or distinctive sentences. Apply the skill when I ask for practitioner tutorials. Ask me to resolve any unclear preference before finalizing the instructions. Sample: [paste your writing].&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h2 id="create-a-style-from-a-description"&gt;
  
  
  Create a style from a description
&lt;/h2&gt;

&lt;p&gt;You can describe the writing behavior directly when you do not have a suitable sample. Give the skill one repeatable task and a clear trigger. Anthropic recommends focused skills with explicit instructions and examples where useful. &lt;a href="https://support.claude.com/en/articles/12512198-how-to-create-custom-skills" rel="nofollow ugc noopener noreferrer"&gt;Custom skill guidance&lt;/a&gt;.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Name the audience and document type.&lt;/li&gt;
&lt;li&gt;Specify tone, structure, and detail level.&lt;/li&gt;
&lt;li&gt;Explain when the guidance should apply.&lt;/li&gt;
&lt;li&gt;Ask Claude to turn that description into a skill.&lt;/li&gt;
&lt;li&gt;Inspect the generated package, upload it, and test it.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2 id="five-custom-style-descriptions-to-paste"&gt;
  
  
  Five custom style descriptions to paste
&lt;/h2&gt;

&lt;p&gt;These original descriptions can form the writing instructions for a skill. Add your own examples and specify the task that should trigger each one. Browse &lt;a href="https://www.promptzone.com/ai-prompts"&gt;AI prompts&lt;/a&gt; for additional tasks to use when testing.&lt;/p&gt;

&lt;h3 id="1-practitioner-tutorial"&gt;
  
  
  1. Practitioner tutorial
&lt;/h3&gt;

&lt;blockquote&gt;
&lt;p&gt;Write for an AI practitioner who understands the goal but has not used this tool. Start with the outcome and prerequisites. Explain each action before introducing the next dependency. Use numbered steps for procedures and short paragraphs for reasoning. Preserve exact names from supplied documentation. Label assumptions and leave unverifiable commands out. End with a practical way to check whether the task succeeded.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h3 id="2-decision-memo"&gt;
  
  
  2. Decision memo
&lt;/h3&gt;

&lt;blockquote&gt;
&lt;p&gt;Write for a team lead deciding what to do next. Open with the recommendation and the condition that would change it. Compare options using the same criteria. Separate evidence from judgment. Include costs or limits only when the supplied sources establish them. Keep the tone direct and measured. Close with the decision needed, who owns it if specified, and the next reversible action.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h3 id="3-patient-technical-explainer"&gt;
  
  
  3. Patient technical explainer
&lt;/h3&gt;

&lt;blockquote&gt;
&lt;p&gt;Explain one concept at a time to a curious beginner. Define unfamiliar terms when they first appear. Use a small example, then explain where the example stops being accurate. Keep necessary technical detail and remove decorative jargon. When I ask a narrow question, answer it before expanding. Use an analogy only if you also connect it back to the actual mechanism.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h3 id="4-research-digest"&gt;
  
  
  4. Research digest
&lt;/h3&gt;

&lt;blockquote&gt;
&lt;p&gt;Write a concise digest for readers evaluating AI research. Begin with the research question. Separate the authors' claims, the evidence they provide, and your interpretation. Retain qualifications and relevant experimental conditions. Cite the supplied primary sources beside the claims they support. Do not manufacture missing results or turn a limited experiment into a general ranking. Finish with questions the evidence leaves open.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h3 id="5-editorial-cleanup"&gt;
  
  
  5. Editorial cleanup
&lt;/h3&gt;

&lt;blockquote&gt;
&lt;p&gt;Edit for clarity while preserving my meaning and degree of certainty. Prefer concrete verbs and familiar words. Remove repeated conclusions, empty praise, and introductions that delay the point. Keep technical names unchanged. Explain substantial meaning changes separately so I can review them. Do not add facts, examples, quotations, or promises unless I supply them. Match the original audience and maintain a consistent voice.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h2 id="styles-vs-project-instructions-vs-profile-preferences-vs-system-prompts"&gt;
  
  
  Styles vs project instructions vs profile preferences vs system prompts
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Control&lt;/th&gt;
&lt;th&gt;Scope and purpose&lt;/th&gt;
&lt;th&gt;Suggested use&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Styles at the November 2024 launch&lt;/td&gt;
&lt;td&gt;Presets and custom writing guidance&lt;/td&gt;
&lt;td&gt;Understand the historical Styles feature&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Skills&lt;/td&gt;
&lt;td&gt;Reusable instructions loaded when relevant&lt;/td&gt;
&lt;td&gt;A recurring tutorial or research-digest format&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Project instructions&lt;/td&gt;
&lt;td&gt;Chats within one project&lt;/td&gt;
&lt;td&gt;A client's audience, terminology, and deliverable requirements&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Profile preferences / Instructions for Claude&lt;/td&gt;
&lt;td&gt;Account-wide guidance&lt;/td&gt;
&lt;td&gt;Your general communication preferences&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;API system prompt&lt;/td&gt;
&lt;td&gt;Context and instructions supplied by an application&lt;/td&gt;
&lt;td&gt;The role and writing rules for an API integration&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Sources: &lt;a href="https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features" rel="nofollow ugc noopener noreferrer"&gt;personalization scope&lt;/a&gt;, &lt;a href="https://support.claude.com/en/articles/12512176-what-are-skills" rel="nofollow ugc noopener noreferrer"&gt;Skills behavior&lt;/a&gt;, &lt;a href="https://platform.claude.com/docs/en/api/messages/create" rel="nofollow ugc noopener noreferrer"&gt;API system parameter&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For project-specific writing, open the project, choose &lt;strong&gt;Set project instructions&lt;/strong&gt;, and save your guidance. Put reusable source material in project knowledge. Anthropic notes that context is not shared between project chats unless added there. &lt;a href="https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects" rel="nofollow ugc noopener noreferrer"&gt;Project setup&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For an API application, supply the writing rules through the documented &lt;code&gt;system&lt;/code&gt; parameter. Treat that as application configuration; do not assume a saved chat preference is your API's configuration. See our &lt;a href="https://www.promptzone.com/llm-api-pricing"&gt;LLM API pricing guide&lt;/a&gt; when planning that separate integration.&lt;/p&gt;

&lt;h2 id="limits-and-troubleshooting"&gt;
  
  
  Limits and troubleshooting
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Skill not used:&lt;/strong&gt; check that it is enabled, make its description identify the task, and request it by name. &lt;strong&gt;Upload unavailable:&lt;/strong&gt; check code execution and organization permissions. &lt;a href="https://support.claude.com/en/articles/12512180-use-skills-in-claude" rel="nofollow ugc noopener noreferrer"&gt;Skills troubleshooting&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Writing still feels wrong:&lt;/strong&gt; compare one response with your reference sample. Identify one concrete problem, such as excessive background or unsuitable headings, and revise that instruction before testing again.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Missing Styles menu:&lt;/strong&gt; use the documented Skills or account-instructions controls.&lt;/p&gt;

&lt;h2 id="faq"&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3 id="where-do-i-find-claude-styles-now"&gt;
  
  
  Where do I find Claude Styles now?
&lt;/h3&gt;

&lt;p&gt;Current help directs reusable tone and format customization to &lt;strong&gt;Customize → Skills&lt;/strong&gt;. For account-wide communication guidance, use &lt;strong&gt;Settings → Instructions for Claude&lt;/strong&gt;.&lt;/p&gt;

&lt;h3 id="can-i-make-claude-write-like-me"&gt;
  
  
  Can I make Claude write like me?
&lt;/h3&gt;

&lt;p&gt;You can supply your writing as an example and ask for reusable guidance. Evaluate the resulting instructions and test them on a different assignment; a sample is not proof of exact voice matching.&lt;/p&gt;

&lt;h3 id="are-claude-writing-skills-free"&gt;
  
  
  Are Claude writing skills free?
&lt;/h3&gt;

&lt;p&gt;The current Skills overview includes the Free plan. Code execution must be enabled, and organization controls apply where relevant. &lt;a href="https://support.claude.com/en/articles/12512176-what-are-skills" rel="nofollow ugc noopener noreferrer"&gt;Availability&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="should-i-use-a-skill-or-project-instructions"&gt;
  
  
  Should I use a skill or project instructions?
&lt;/h3&gt;

&lt;p&gt;Use a skill for a repeatable writing method across tasks. Use project instructions when the guidance belongs to one project's audience, context, or requirements.&lt;/p&gt;

&lt;h2 id="sources"&gt;
  
  
  Sources
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features" rel="nofollow ugc noopener noreferrer"&gt;Current Claude personalization guide&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.anthropic.com/news/styles" rel="nofollow ugc noopener noreferrer"&gt;Anthropic's Styles launch announcement, November 26, 2024 (indexed historical text; live page returns 404)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://support.claude.com/en/articles/12512176-what-are-skills" rel="nofollow ugc noopener noreferrer"&gt;What are Skills?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://support.claude.com/en/articles/12512180-use-skills-in-claude" rel="nofollow ugc noopener noreferrer"&gt;Use Skills in Claude&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://support.claude.com/en/articles/12512198-how-to-create-custom-skills" rel="nofollow ugc noopener noreferrer"&gt;How to create custom Skills&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://academy.claude.com/tutorials/teach-claude-your-way-of-working-using-skills" rel="nofollow ugc noopener noreferrer"&gt;Teach Claude your way of working using Skills&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects" rel="nofollow ugc noopener noreferrer"&gt;Create and manage projects&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://platform.claude.com/docs/en/api/messages/create" rel="nofollow ugc noopener noreferrer"&gt;Messages API reference: system instructions&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>claude</category>
      <category>ai</category>
      <category>writing</category>
      <category>prompts</category>
    </item>
    <item>
      <title>Top 8 AI Predictions for 2025: Insights from the State of AI Report 2024</title>
      <dc:creator>celine</dc:creator>
      <pubDate>Tue, 22 Oct 2024 06:31:22 +0000</pubDate>
      <link>https://www.promptzone.com/celine/top-8-ai-predictions-for-2025-insights-from-the-state-of-ai-report-2024-k2b</link>
      <guid>https://www.promptzone.com/celine/top-8-ai-predictions-for-2025-insights-from-the-state-of-ai-report-2024-k2b</guid>
      <description>&lt;p&gt;The State of AI Report 2024 has recently been released, offering a comprehensive look at the current landscape and future trends in artificial intelligence. Among its many insights, the report includes predictions for 2025 that have caught our attention. Let's explore these forecasts and provide our honest reactions to each.&lt;/p&gt;

&lt;h2 id="1-national-security-review-triggered-by-sovereign-wealth-investment"&gt;
  
  
  1. National Security Review Triggered by Sovereign Wealth Investment
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; A significant sovereign wealth investment in a US AI lab will prompt a national security review.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; This scenario is already unfolding to some extent. As AI becomes increasingly critical to national interests, we expect heightened scrutiny of foreign investments in this sector.&lt;/p&gt;

&lt;h2 id="2-viral-app-created-by-ai-novice"&gt;
  
  
  2. Viral App Created by AI Novice
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; An application developed by an AI beginner will go viral.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; While tools like NotebookLM have shown promise, we haven't seen a true breakout app from a novice yet. This prediction remains plausible but unfulfilled.&lt;/p&gt;

&lt;h2 id="3-improved-data-collection-practices-by-frontier-labs"&gt;
  
  
  3. Improved Data Collection Practices by Frontier Labs
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; Leading AI research labs will adopt more rigorous data collection practices.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; Some progress has been made, particularly in response to EU regulations. However, there's still significant room for improvement in transparency and ethical data practices.&lt;/p&gt;

&lt;h2 id="4-opensource-ai-surpasses-gpt4"&gt;
  
  
  4. Open-Source AI Surpasses GPT-4
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; An open-source alternative will outperform GPT-4.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; Numerous projects are working towards this goal, but achieving GPT-4 level performance remains a significant challenge. We're watching this space closely.&lt;/p&gt;

&lt;h2 id="5-nvidias-market-dominance-continues"&gt;
  
  
  5. NVIDIA's Market Dominance Continues
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; Competitors will struggle to challenge NVIDIA's position in the AI hardware market.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; NVIDIA's lead seems secure for now, but rapid advancements in AI hardware could lead to surprises. NVIDIA's biggest challenge may come from within, as they navigate scaling and market expectations.&lt;/p&gt;

&lt;h2 id="6-cooling-interest-in-humanoid-robotics"&gt;
  
  
  6. Cooling Interest in Humanoid Robotics
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; Enthusiasm for humanoid robotics will decrease.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; We disagree with this prediction. We anticipate continued interest in robotics, with practical applications driving innovation in both hardware and software.&lt;/p&gt;

&lt;h2 id="7-success-of-apples-ondevice-ai"&gt;
  
  
  7. Success of Apple's On-Device AI
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; Apple's efforts in on-device AI will yield significant results.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; With the imminent release of Apple Intelligence Phase 1, we'll soon have concrete evidence of Apple's progress in this area. The potential for privacy-preserving, on-device AI is substantial.&lt;/p&gt;

&lt;h2 id="8-breakthrough-in-aipowered-gaming"&gt;
  
  
  8. Breakthrough in AI-Powered Gaming
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prediction:&lt;/strong&gt; A video game leveraging generative AI will achieve significant success.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Our Take:&lt;/strong&gt; While this seems ambitious for the near term, the rapid pace of AI development makes it a possibility. We're excited to see how AI integration evolves in the gaming industry.&lt;/p&gt;

&lt;h2 id="conclusion"&gt;
  
  
  Conclusion
&lt;/h2&gt;

&lt;p&gt;The State of AI Report 2024 offers fascinating insights into the future of artificial intelligence. While some predictions seem more likely than others, the dynamic nature of AI technology means surprises are always possible. As we move towards 2025, we'll be keenly observing how these predictions unfold and what new developments emerge in the world of AI.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>news</category>
    </item>
  </channel>
</rss>
