<?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: Thomas</title>
    <description>The latest articles on PromptZone - AI Prompts, Guides and Tools for Builders by Thomas (@stabletom).</description>
    <link>https://www.promptzone.com/stabletom</link>
    <image>
      <url>https://promptzone-community.s3.amazonaws.com/uploads/user/profile_image/815/1f96050e-72e4-4f46-bff8-4ebb340179be.jpg</url>
      <title>PromptZone - AI Prompts, Guides and Tools for Builders: Thomas</title>
      <link>https://www.promptzone.com/stabletom</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://www.promptzone.com/feed/stabletom"/>
    <language>en</language>
    <item>
      <title>Creating free images with Flux [pro]</title>
      <dc:creator>Thomas</dc:creator>
      <pubDate>Fri, 23 Aug 2024 19:16:23 +0000</pubDate>
      <link>https://www.promptzone.com/stabletom/creating-free-images-with-flux-pro-52o9</link>
      <guid>https://www.promptzone.com/stabletom/creating-free-images-with-flux-pro-52o9</guid>
      <description>&lt;h2 id="flux-pro-free"&gt;
  
  
  Flux [pro] Free
&lt;/h2&gt;

&lt;h3 id="what-is-it"&gt;
  
  
  What is it?
&lt;/h3&gt;

&lt;p&gt;Flux [pro] is the most advanced artificial intelligence model for image generation in the [Flux] suite from Black Forest Labs.&lt;/p&gt;

&lt;p&gt;This professional version offers enhanced capabilities for creating very high quality visuals of a level equivalent, if not superior, to private models such as &lt;a href="https://www.promptzone.com/deepa_kowalski/ai-image-generators-2026-vheer-visualgpt-fooocus-comfyui-midjourney-more-compared-2i44"&gt;Midjourney&lt;/a&gt; or DALL-E.&lt;/p&gt;

&lt;h3 id="using-flux-pro"&gt;
  
  
  Using Flux [pro]
&lt;/h3&gt;

&lt;p&gt;At the time of its launch, Flux [pro] was only available via paid access from Black Forest Labs partners. Today, however, it's available free of charge from a new site: fluxpro.art.&lt;/p&gt;

&lt;p&gt;Here's how to use this online generator to create stunning, fantastic images with Flux [pro] :&lt;/p&gt;

&lt;h4 id="generate-an-image-on-fluxproart"&gt;
  
  
  Generate an image on fluxpro.art
&lt;/h4&gt;

&lt;ol&gt;
&lt;li&gt;&lt;p&gt;Go to the [fluxpro.art] website (&lt;a href="http://fluxpro.art" rel="nofollow ugc noopener noreferrer"&gt;http://fluxpro.art&lt;/a&gt;) and log in by clicking on &lt;em&gt;Login&lt;/em&gt; in the top right-hand corner (secure connections via a Google account). &lt;br&gt;
&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/u2dzt5r4a7p15tlfhey1.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/u2dzt5r4a7p15tlfhey1.png" alt="fluxpro.art homepage"&gt;&lt;/a&gt;&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Then open the generator by clicking on &lt;a href="https://fluxpro.art/create" rel="nofollow ugc noopener noreferrer"&gt;&lt;em&gt;Create&lt;/em&gt;&lt;/a&gt; in the menu at the top of the page.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Enter a prompt in the text field.&lt;br&gt;
&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/amehhxlzcs108pg0nc9u.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/amehhxlzcs108pg0nc9u.png" alt="fluxpro interface"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;For inspiration, have a look at these &lt;a href="https://www.stablediffusion.blog/prompts-flux" rel="nofollow ugc noopener noreferrer"&gt;Flux prompts&lt;/a&gt; shared by French-speaking designers.&lt;/em&gt;&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Finally, click on the &lt;em&gt;Generate&lt;/em&gt; button to start the generation process: Flux [pro] works its magic, and in just a few seconds, 4 magnificent images appear.&lt;br&gt;
&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/81xs5yvjccu5rs9cr00o.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/81xs5yvjccu5rs9cr00o.png" alt="sample generation"&gt;&lt;/a&gt;&lt;/p&gt;&lt;/li&gt;
&lt;/ol&gt;

&lt;h4 id="flux-pro-parameters"&gt;
  
  
  Flux [pro] parameters
&lt;/h4&gt;

&lt;p&gt;Fluxpro.art lets you modify several generation parameters to make the most of the power of Flux [pro]:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;&lt;em&gt;Aspect Ratio&lt;/em&gt;&lt;/strong&gt; designates the image format and determines its dimensions:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Format&lt;/th&gt;
&lt;th&gt;Orientation&lt;/th&gt;
&lt;th&gt;Width&lt;/th&gt;
&lt;th&gt;Height&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;1:1&lt;/td&gt;
&lt;td&gt;Square&lt;/td&gt;
&lt;td&gt;1024px&lt;/td&gt;
&lt;td&gt;1024px&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;16:9&lt;/td&gt;
&lt;td&gt;Landscape&lt;/td&gt;
&lt;td&gt;1344px&lt;/td&gt;
&lt;td&gt;768px&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;9:16 Portrait&lt;/td&gt;
&lt;td&gt;768px&lt;/td&gt;
&lt;td&gt;1344px&lt;/td&gt;
&lt;td&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;3:2&lt;/td&gt;
&lt;td&gt;Landscape&lt;/td&gt;
&lt;td&gt;1216px&lt;/td&gt;
&lt;td&gt;832px&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;2:3 Portrait&lt;/td&gt;
&lt;td&gt;832px&lt;/td&gt;
&lt;td&gt;1216px&lt;/td&gt;
&lt;td&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;5:4&lt;/td&gt;
&lt;td&gt;Landscape&lt;/td&gt;
&lt;td&gt;1088px&lt;/td&gt;
&lt;td&gt;896px&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;4:5 Portrait&lt;/td&gt;
&lt;td&gt;896px&lt;/td&gt;
&lt;td&gt;1088px&lt;/td&gt;
&lt;td&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;4:3&lt;/td&gt;
&lt;td&gt;Landscape&lt;/td&gt;
&lt;td&gt;1152px&lt;/td&gt;
&lt;td&gt;864px&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;3:4&lt;/td&gt;
&lt;td&gt;Portrait&lt;/td&gt;
&lt;td&gt;864px&lt;/td&gt;
&lt;td&gt;1152px&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;&lt;em&gt;Safety&lt;/em&gt;&lt;/strong&gt; indicates the level of image moderation&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;Safe&lt;/em&gt;&lt;/strong&gt;: Suitable for work. Nothing inappropriate.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;Mid&lt;/em&gt;&lt;/strong&gt; : Revealing clothing, violence or slight bloody elements.&lt;/li&gt;
&lt;li&gt;Mature***: Adult themes and situations, partial nudity, graphic violence or death.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;Explicit&lt;/em&gt;&lt;/strong&gt;: Disturbing sexual or graphic content. Graphic nudity, adult objects or contexts.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;&lt;em&gt;Steps&lt;/em&gt;&lt;/strong&gt; indicates the number of iteration steps in image generation. A higher number generally generates a more detailed image - but lengthens the generation time and may have undesirable effects on image quality and rendering.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;&lt;em&gt;Guidance&lt;/em&gt;&lt;/strong&gt; corresponds to the degree to which the template respects the text. The higher it is, the more the template will try to respect the prompt and the less creative it will be.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;&lt;em&gt;Seed&lt;/em&gt;&lt;/strong&gt; determines the random starting state of the image at the start of the generation process. The same seed with exactly the same parameters will produce the same images.&lt;/p&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;h3 id="explore-and-be-inspired"&gt;
  
  
  Explore and be inspired
&lt;/h3&gt;

&lt;p&gt;In addition to the Flux [pro] image generator, the fluxpro.art website also features an image gallery containing all the images generated by its users, with their prompts visible. With over 2 million images to choose from, there's plenty of scope for finding thousands of ideas and discovering the talents of AI.&lt;/p&gt;

&lt;p&gt;Clicking on an image brings up its prompt and 5 action buttons:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/e6e7sm3hu040iyluvkqt.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/e6e7sm3hu040iyluvkqt.png" alt="showing an image"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;In order, the buttons are :&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;Copy prompt&lt;/em&gt;&lt;/strong&gt; copies the prompt, which you can then use as you wish.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;External Link&lt;/em&gt;&lt;/strong&gt; opens a page with the image and details - ideal for sharing your creations on social networks.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;Report&lt;/em&gt;&lt;/strong&gt; allows you to report content for moderation.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;Bookmark&lt;/em&gt;&lt;/strong&gt; bookmarks the image, accessible via the &lt;em&gt;Bookmarks&lt;/em&gt; link in the main menu.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;em&gt;Like&lt;/em&gt;&lt;/strong&gt; allows you to like an image.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Above the prompt, you'll also see the designer's name (click on it to see his profile and other images) and a Follow button to follow him on fluxpro.art.&lt;/p&gt;

&lt;h2 id="now-its-your-turn"&gt;
  
  
  Now it's your turn
&lt;/h2&gt;

&lt;p&gt;Now you know everything you need to know to create free images with Flux [pro] and share them with the Fluxpro.art community - all you have to do is create!&lt;/p&gt;

&lt;p&gt;Don't forget to &lt;a href="https://fluxpro.art/users/cm03o8pt100d5wciavo0f7kuf" rel="nofollow ugc noopener noreferrer"&gt;take a look at my profile&lt;/a&gt; and follow me... I'll be sharing some cool prompts there regularly.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>flux</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>Realistic Photos with Flux: 2026 Prompt Guide</title>
      <dc:creator>Thomas</dc:creator>
      <pubDate>Fri, 16 Aug 2024 11:20:46 +0000</pubDate>
      <link>https://www.promptzone.com/stabletom/realistic-photos-with-flux-57aa</link>
      <guid>https://www.promptzone.com/stabletom/realistic-photos-with-flux-57aa</guid>
      <description>&lt;p&gt;For flux realistic photos, describe a believable subject, lighting setup, camera viewpoint, and surface texture. Use FLUX.1-dev for the XLabs realism LoRA; choose a dedicated Kontext or FLUX.2 workflow for image editing and newer models.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/0w57d49pos2lblwknw20.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/0w57d49pos2lblwknw20.png" alt="Follow @StableTom"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2 id="flux-realismlora"&gt;
  
  
  Flux RealismLora
&lt;/h2&gt;

&lt;p&gt;The &lt;a href="https://huggingface.co/XLabs-AI/flux-RealismLora" rel="ugc noopener noreferrer"&gt;XLabs realism LoRA&lt;/a&gt; is an adapter trained for photorealism on &lt;strong&gt;FLUX.1-dev&lt;/strong&gt;. Its downloadable file is &lt;code&gt;lora.safetensors&lt;/code&gt;, and its card specifies the FLUX.1-dev Non-Commercial License.&lt;/p&gt;

&lt;p&gt;Load the adapter alongside its base model. Do not assume that a file described as “flux realism lora” works with schnell, Kontext, or FLUX.2: the XLabs card identifies dev specifically.&lt;/p&gt;

&lt;p&gt;For background on the original family, see &lt;a href="https://www.promptzone.com/stabletom/flux-the-new-prodigy-model-2ede"&gt;Flux: the new prodigy model&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/g07934glwavy6hk3s6ni.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/g07934glwavy6hk3s6ni.png" alt="man holding a sign that say "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2 id="which-flux-model-should-you-use"&gt;
  
  
  Which FLUX model should you use?
&lt;/h2&gt;

&lt;p&gt;Pick the model before downloading adapters. These are distinct workflows with different weight licenses, input requirements, and controls.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Model&lt;/th&gt;
&lt;th&gt;Relevant use&lt;/th&gt;
&lt;th&gt;Weight license or access&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-dev" rel="ugc noopener noreferrer"&gt;FLUX.1-dev&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Text-to-image; base for XLabs realism&lt;/td&gt;
&lt;td&gt;FLUX.1-dev Non-Commercial License&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-schnell" rel="ugc noopener noreferrer"&gt;FLUX.1-schnell&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Fast text-to-image drafts&lt;/td&gt;
&lt;td&gt;Apache 2.0&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev" rel="ugc noopener noreferrer"&gt;FLUX.1 Kontext dev&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Edit an image with text instructions&lt;/td&gt;
&lt;td&gt;FLUX.1-dev Non-Commercial License&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.2-dev" rel="ugc noopener noreferrer"&gt;FLUX.2-dev&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Generate, edit, and combine reference images locally&lt;/td&gt;
&lt;td&gt;FLUX Non-Commercial License&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.2-klein-4B" rel="ugc noopener noreferrer"&gt;FLUX.2 klein 4B&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Compact generation and editing&lt;/td&gt;
&lt;td&gt;Apache 2.0&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;FLUX.2 klein 9B&lt;/td&gt;
&lt;td&gt;Generation and editing with open weights&lt;/td&gt;
&lt;td&gt;FLUX Non-Commercial License&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;FLUX.2 pro, max, flex&lt;/td&gt;
&lt;td&gt;Hosted generation and editing&lt;/td&gt;
&lt;td&gt;API access under service terms&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;BFL also offers undistilled klein Base variants for training and customization. Their licenses follow the same 4B/9B split. See the &lt;a href="https://docs.bfl.ai/flux_2/flux2_overview" rel="ugc noopener noreferrer"&gt;FLUX.2 family overview&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The dev cards distinguish model use from output use: commercial outputs can be permitted under their terms even though the weights have non-commercial licenses. Check the exact license for your deployment.&lt;/p&gt;

&lt;p&gt;BFL now also presents FLUX 3 on &lt;a href="https://bfl.ai/" rel="ugc noopener noreferrer"&gt;its website&lt;/a&gt;, with video capabilities documented in its &lt;a href="https://docs.bfl.ai/release-notes" rel="ugc noopener noreferrer"&gt;release notes&lt;/a&gt;. For still-image generation, BFL's &lt;a href="https://docs.bfl.ai/flux_2/flux2_text_to_image" rel="ugc noopener noreferrer"&gt;text-to-image guide&lt;/a&gt; recommends FLUX.2.&lt;/p&gt;

&lt;h2 id="flux-on-falai"&gt;
  
  
  Flux on fal.ai
&lt;/h2&gt;

&lt;p&gt;The &lt;a href="https://fal.ai/models/fal-ai/flux-realism" rel="ugc noopener noreferrer"&gt;dedicated fal realism page&lt;/a&gt; is marked &lt;strong&gt;Private&lt;/strong&gt;. Access is account-dependent; do not build your workflow around an assumption that every signed-in user can run it.&lt;/p&gt;

&lt;p&gt;The following screenshots illustrate the hosted workflow. Follow the named model and fields when the interface layout differs.&lt;/p&gt;

&lt;h3 id="stepbystep-guide"&gt;
  
  
  Step-by-step guide
&lt;/h3&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Open the model playground and sign in.&lt;/strong&gt; Check access before preparing a batch. No particular third-party login method is required by these instructions.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/otl0odrqp07e89cb4msq.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/otl0odrqp07e89cb4msq.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Choose the compatible endpoint.&lt;/strong&gt; If the dedicated page is unavailable, use &lt;a href="https://fal.ai/models/fal-ai/flux-lora/api" rel="ugc noopener noreferrer"&gt;fal's FLUX.1-dev LoRA endpoint&lt;/a&gt;. Its &lt;code&gt;loras&lt;/code&gt; input accepts a weight URL in &lt;code&gt;path&lt;/code&gt; and an adapter &lt;code&gt;scale&lt;/code&gt;.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/dazmh1rgz3v7m12e5uvp.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/dazmh1rgz3v7m12e5uvp.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Enter your prompt.&lt;/strong&gt; For the general LoRA endpoint, use this adapter entry alongside the required &lt;code&gt;prompt&lt;/code&gt;:
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight json"&gt;&lt;code&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="w"&gt;
  &lt;/span&gt;&lt;span class="nl"&gt;"loras"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="p"&gt;[{&lt;/span&gt;&lt;span class="w"&gt;
    &lt;/span&gt;&lt;span class="nl"&gt;"path"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="s2"&gt;"https://huggingface.co/XLabs-AI/flux-RealismLora/resolve/main/lora.safetensors"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;&lt;span class="w"&gt;
    &lt;/span&gt;&lt;span class="nl"&gt;"scale"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="w"&gt;
  &lt;/span&gt;&lt;span class="p"&gt;}]&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/9wb043m2qjfupewiqiv9.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/9wb043m2qjfupewiqiv9.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Run the generation and save the result.&lt;/strong&gt; Inspect the image before repeating the request; adjust the description if the composition misses your intention.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/god90bsgdf1mntg15zw6.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/god90bsgdf1mntg15zw6.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Open additional settings.&lt;/strong&gt; Record the seed, dimensions, steps, guidance, and adapter strength alongside the prompt. The &lt;a href="https://fal.ai/models/fal-ai/flux-realism/api" rel="ugc noopener noreferrer"&gt;realism API schema&lt;/a&gt; documents its controls and defaults.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/smbsg1vjgpu17pfhu7zj.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/smbsg1vjgpu17pfhu7zj.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Replicate alternative:&lt;/strong&gt; &lt;a href="https://replicate.com/xlabs-ai/flux-dev-realism" rel="ugc noopener noreferrer"&gt;XLabs' flux-dev-realism&lt;/a&gt; explicitly combines FLUX.1-dev with the XLabs adapter. Open its Playground, enter a prompt, review the available inputs, and run a prediction.&lt;/p&gt;

&lt;p&gt;Check the selected provider's displayed billing terms before generating. Treat free-credit offers as account-specific rather than part of the workflow.&lt;/p&gt;

&lt;h2 id="run-the-realism-lora-locally-in-comfyui"&gt;
  
  
  Run the realism LoRA locally in ComfyUI
&lt;/h2&gt;

&lt;p&gt;Start with the &lt;a href="https://comfyanonymous.github.io/ComfyUI_examples/flux/" rel="ugc noopener noreferrer"&gt;official ComfyUI FLUX examples&lt;/a&gt;. Drag the dev example image into ComfyUI to load its workflow, then install its model files:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;File&lt;/th&gt;
&lt;th&gt;Folder inside ComfyUI&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;flux1-dev.safetensors&lt;/code&gt;&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;&lt;code&gt;clip_l.safetensors&lt;/code&gt;&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;&lt;code&gt;t5xxl_fp16.safetensors&lt;/code&gt;&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;&lt;code&gt;ae.safetensors&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;&lt;code&gt;models/vae/&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;For lower memory use, the examples document &lt;code&gt;t5xxl_fp8_e4m3fn_scaled.safetensors&lt;/code&gt; and the diffusion loader's &lt;code&gt;weight_dtype&lt;/code&gt; FP8 option. The all-in-one FP8 checkpoint instead goes in &lt;code&gt;models/checkpoints/&lt;/code&gt; and uses Load Checkpoint.&lt;/p&gt;

&lt;p&gt;FP8 can reduce image quality. Compare it with the full-precision workflow when your hardware permits, rather than treating quantization as a realism improvement.&lt;/p&gt;

&lt;p&gt;For the adapter author's workflow, install &lt;a href="https://github.com/XLabs-AI/x-flux-comfyui" rel="ugc noopener noreferrer"&gt;x-flux-comfyui&lt;/a&gt; using ComfyUI's Python environment. From the ComfyUI directory:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/XLabs-AI/x-flux-comfyui custom_nodes/x-flux-comfyui
&lt;span class="nb"&gt;cd &lt;/span&gt;custom_nodes/x-flux-comfyui
python setup.py
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Restart ComfyUI. Put the XLabs &lt;code&gt;lora.safetensors&lt;/code&gt; in &lt;code&gt;models/xlabs/loras/&lt;/code&gt;, as specified by the extension README, and load an example from its &lt;code&gt;workflows&lt;/code&gt; folder.&lt;/p&gt;

&lt;p&gt;The author's &lt;a href="https://github.com/XLabs-AI/x-flux-comfyui/blob/main/Guide.md" rel="ugc noopener noreferrer"&gt;node guide&lt;/a&gt; uses &lt;strong&gt;Load Flux LoRA&lt;/strong&gt; to modify the incoming model. Select the adapter there and send its model output into the sampler.&lt;/p&gt;

&lt;p&gt;For GGUF, install &lt;a href="https://github.com/city96/ComfyUI-GGUF" rel="ugc noopener noreferrer"&gt;ComfyUI-GGUF&lt;/a&gt; in the same environment. From the ComfyUI directory:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/city96/ComfyUI-GGUF custom_nodes/ComfyUI-GGUF
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;--upgrade&lt;/span&gt; gguf
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Restart, put a compatible FLUX.1-dev &lt;code&gt;.gguf&lt;/code&gt; in &lt;code&gt;models/unet/&lt;/code&gt;, and replace Load Diffusion Model with &lt;strong&gt;Unet Loader (GGUF)&lt;/strong&gt;. Keep the text encoders and VAE required by the workflow.&lt;/p&gt;

&lt;p&gt;The GGUF project calls LoRA loading experimental. Validate the adapter on a simple workflow before adding other custom nodes. For general workflow navigation, use the &lt;a href="https://www.promptzone.com/tomas_novak/comfyui-2026-the-complete-guide-to-power-user-ai-image-generation-1g17"&gt;ComfyUI guide&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="settings-that-matter"&gt;
  
  
  Settings that matter
&lt;/h2&gt;

&lt;p&gt;These are documented examples or endpoint defaults, not a universal realism recipe. A setting belongs to its model and implementation; hosted controls do not necessarily map directly onto ComfyUI node labels.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Model or endpoint&lt;/th&gt;
&lt;th&gt;Guidance&lt;/th&gt;
&lt;th&gt;Steps&lt;/th&gt;
&lt;th&gt;Resolution&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;FLUX.1-dev, BFL card example&lt;/td&gt;
&lt;td&gt;3.5&lt;/td&gt;
&lt;td&gt;50&lt;/td&gt;
&lt;td&gt;1024 × 1024&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;FLUX.1-schnell&lt;/td&gt;
&lt;td&gt;0.0 in Diffusers&lt;/td&gt;
&lt;td&gt;4; model supports 1–4&lt;/td&gt;
&lt;td&gt;1024 × 1024 in the ComfyUI template&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;fal FLUX realism&lt;/td&gt;
&lt;td&gt;3.5 default&lt;/td&gt;
&lt;td&gt;28 default&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;landscape_4_3&lt;/code&gt; default; custom dimensions supported&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Kontext dev, ComfyUI template&lt;/td&gt;
&lt;td&gt;2.5&lt;/td&gt;
&lt;td&gt;20&lt;/td&gt;
&lt;td&gt;Input image through FluxKontextImageScale&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;FLUX.2-dev&lt;/td&gt;
&lt;td&gt;4&lt;/td&gt;
&lt;td&gt;20 in ComfyUI; 50 in BFL's example&lt;/td&gt;
&lt;td&gt;1024 × 1024 in the ComfyUI template&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;FLUX.2 klein 4B, distilled card example&lt;/td&gt;
&lt;td&gt;1.0&lt;/td&gt;
&lt;td&gt;4&lt;/td&gt;
&lt;td&gt;1024 × 1024&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;FLUX.2 pro/max API&lt;/td&gt;
&lt;td&gt;Managed by endpoint&lt;/td&gt;
&lt;td&gt;Managed by endpoint&lt;/td&gt;
&lt;td&gt;BFL's generation example uses 1024 × 1024&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The native examples are the &lt;a href="https://github.com/Comfy-Org/workflow_templates/blob/main/templates/flux_schnell.json" rel="ugc noopener noreferrer"&gt;schnell&lt;/a&gt;, &lt;a href="https://github.com/Comfy-Org/workflow_templates/blob/main/templates/flux_kontext_dev_basic.json" rel="ugc noopener noreferrer"&gt;Kontext&lt;/a&gt;, and &lt;a href="https://github.com/Comfy-Org/workflow_templates/blob/main/templates/image_flux2_text_to_image.json" rel="ugc noopener noreferrer"&gt;FLUX.2&lt;/a&gt; templates. Preserve their node settings when comparing models.&lt;/p&gt;

&lt;p&gt;For the ComfyUI all-in-one FLUX dev FP8 example, set the sampler's &lt;strong&gt;CFG to 1.0&lt;/strong&gt;, as its instructions require. Do not copy the card's guidance value into that CFG field.&lt;/p&gt;

&lt;h2 id="fluxrealism-prompts"&gt;
  
  
  Flux-Realism prompts
&lt;/h2&gt;

&lt;p&gt;Build a scene from these seven ingredients, then remove anything that does not change the intended picture:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Subject:&lt;/strong&gt; a middle-aged reader with salt-and-pepper hair.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Action:&lt;/strong&gt; lifting a coffee cup while reading.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Location:&lt;/strong&gt; a café beside a street-facing window.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Details:&lt;/strong&gt; a creased shirt, worn table, and folded newspaper.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ambiance:&lt;/strong&gt; a quiet morning with subdued colors.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Purpose:&lt;/strong&gt; a candid moment during a daily routine.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Style:&lt;/strong&gt; documentary or amateur photography.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;This follows BFL's &lt;a href="https://docs.bfl.ai/guides/prompting_unified_building" rel="ugc noopener noreferrer"&gt;prompt-building guidance&lt;/a&gt;: establish the subject and action, then add relevant visual direction. Avoid padding a precise scene with unrelated quality adjectives.&lt;/p&gt;

&lt;h2 id="prompting-for-realism"&gt;
  
  
  Prompting for realism
&lt;/h2&gt;

&lt;p&gt;BFL's &lt;a href="https://docs.bfl.ai/guides/usecases_t2i_photorealistic" rel="ugc noopener noreferrer"&gt;photorealism guide&lt;/a&gt; recommends describing lens, framing, lighting, and physical texture. Ask for visible pores, fabric wear, or uneven surfaces where those details suit the scene.&lt;/p&gt;

&lt;p&gt;Use camera language to specify an intended appearance. These are prompt cues, not a guarantee that the model will simulate an actual camera's optics.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Intended appearance&lt;/th&gt;
&lt;th&gt;Vocabulary to try&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Environmental portrait&lt;/td&gt;
&lt;td&gt;35mm lens, eye-level framing, visible surroundings&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Closer portrait&lt;/td&gt;
&lt;td&gt;85mm lens, f/2.8, softly blurred background&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Broad scene&lt;/td&gt;
&lt;td&gt;24mm lens, deep depth of field&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Gentle illumination&lt;/td&gt;
&lt;td&gt;Diffused window light, soft shadows&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Dramatic portrait&lt;/td&gt;
&lt;td&gt;Rembrandt lighting, restrained background&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Studio subject&lt;/td&gt;
&lt;td&gt;Softbox illumination, controlled highlights&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;These lens and lighting categories come from BFL's &lt;a href="https://docs.bfl.ai/guides/prompting_unified_reference" rel="ugc noopener noreferrer"&gt;prompt reference&lt;/a&gt;. Choose a coherent combination: a portrait with a soft background asks for a different look from an evenly sharp interior.&lt;/p&gt;

&lt;p&gt;For FLUX.2, use positive descriptions instead of relying on negative prompts. Its &lt;a href="https://docs.bfl.ai/guides/prompting_unified_technical" rel="ugc noopener noreferrer"&gt;technical guide&lt;/a&gt; recommends describing what should appear: “an empty tabletop” makes the desired scene explicit.&lt;/p&gt;

&lt;h2 id="a-few-prompts"&gt;
  
  
  A few prompts
&lt;/h2&gt;

&lt;p&gt;Use these original prompt variations as starting points. Keep the subject fixed while testing a lighting or framing change, so you can judge which instruction improves your result.&lt;/p&gt;

&lt;h3 id="café-portrait"&gt;
  
  
  Café portrait
&lt;/h3&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/hxkemz989l3ugorcio8i.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/hxkemz989l3ugorcio8i.png" alt="man drinking coffee"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;A candid photograph of a middle-aged man with salt-and-pepper hair, a white shirt, and navy blazer, sipping coffee beside an open newspaper. A worn café table and hanging plants frame him.&lt;/p&gt;

&lt;p&gt;Eye-level view, 35mm lens, diffused window light, visible shirt creases, natural skin texture, subdued morning colors. Quiet concentration rather than a posed smile.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h3 id="working-chef"&gt;
  
  
  Working chef
&lt;/h3&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/unjz020tntkx6igx1mk3.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/unjz020tntkx6igx1mk3.png" alt="female chef"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;An editorial photograph of a female chef in her forties, curly hair tucked under a white hat, carefully placing a red pepper slice on a dish. Copper pans and steel counters surround her.&lt;/p&gt;

&lt;p&gt;Close framing, soft side light, wisps of steam, textured jacket fabric, and restrained reflections. Keep her expression focused and the food's surface clearly visible.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h3 id="library-documentary-scene"&gt;
  
  
  Library documentary scene
&lt;/h3&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/z9osygqq6wwy2iykc0gw.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/z9osygqq6wwy2iykc0gw.png" alt="cross-legged student"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;A documentary photograph of an Asian university student in a grey hoodie, sitting cross-legged among open books with a laptop on his knees. Tall shelves and distant students establish the library setting.&lt;/p&gt;

&lt;p&gt;Wide-angle framing, warm desk lamps, worn book covers, rumpled clothing, and a half-finished coffee. An absorbed study session with ordinary, imperfect surroundings.&lt;/p&gt;
&lt;/blockquote&gt;

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

&lt;h3 id="do-i-need-a-realism-lora-for-realistic-flux-photos"&gt;
  
  
  Do I need a realism LoRA for realistic FLUX photos?
&lt;/h3&gt;

&lt;p&gt;No. BFL documents photorealistic prompting for FLUX.2 without an adapter. XLabs provides a specific FLUX.1-dev option; compare it against the base model using the same scene.&lt;/p&gt;

&lt;h3 id="does-xlabs-realism-lora-work-with-flux2"&gt;
  
  
  Does XLabs realism LoRA work with FLUX.2?
&lt;/h3&gt;

&lt;p&gt;Its model card specifies FLUX.1-dev. Use that supported pairing; a shared FLUX name does not establish adapter compatibility.&lt;/p&gt;

&lt;h3 id="is-fals-flux-realism-endpoint-public"&gt;
  
  
  Is fal's FLUX realism endpoint public?
&lt;/h3&gt;

&lt;p&gt;Its page is marked Private. Check your account's access, use fal's general FLUX LoRA endpoint, or try the XLabs-hosted model on Replicate.&lt;/p&gt;

&lt;h3 id="can-i-use-flux-commercially"&gt;
  
  
  Can I use FLUX commercially?
&lt;/h3&gt;

&lt;p&gt;Schnell and klein 4B carry Apache 2.0 licenses. Dev-family weights have their stated non-commercial terms, with separate provisions for outputs. Hosted use follows the provider's terms.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://huggingface.co/XLabs-AI/flux-RealismLora" rel="ugc noopener noreferrer"&gt;XLabs realism LoRA card&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-dev" rel="ugc noopener noreferrer"&gt;FLUX.1-dev card&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-schnell" rel="ugc noopener noreferrer"&gt;FLUX.1-schnell card&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev" rel="ugc noopener noreferrer"&gt;Kontext dev card&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.2-dev" rel="ugc noopener noreferrer"&gt;FLUX.2-dev card&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.2-klein-4B" rel="ugc noopener noreferrer"&gt;Klein 4B card&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.bfl.ai/flux_2/flux2_overview" rel="ugc noopener noreferrer"&gt;FLUX.2 family overview&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://bfl.ai/" rel="ugc noopener noreferrer"&gt;Black Forest Labs&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.bfl.ai/release-notes" rel="ugc noopener noreferrer"&gt;BFL release notes&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.bfl.ai/flux_2/flux2_text_to_image" rel="ugc noopener noreferrer"&gt;FLUX.2 generation guide&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://fal.ai/models/fal-ai/flux-realism" rel="ugc noopener noreferrer"&gt;fal realism playground&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://fal.ai/models/fal-ai/flux-realism/api" rel="ugc noopener noreferrer"&gt;fal realism schema&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://fal.ai/models/fal-ai/flux-lora/api" rel="ugc noopener noreferrer"&gt;fal FLUX LoRA schema&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://replicate.com/xlabs-ai/flux-dev-realism" rel="ugc noopener noreferrer"&gt;Replicate XLabs realism&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://comfyanonymous.github.io/ComfyUI_examples/flux/" rel="ugc noopener noreferrer"&gt;ComfyUI FLUX examples&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/Comfy-Org/workflow_templates/blob/main/templates/flux_schnell.json" rel="ugc noopener noreferrer"&gt;ComfyUI schnell template&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/Comfy-Org/workflow_templates/blob/main/templates/flux_kontext_dev_basic.json" rel="ugc noopener noreferrer"&gt;ComfyUI Kontext template&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/Comfy-Org/workflow_templates/blob/main/templates/image_flux2_text_to_image.json" rel="ugc noopener noreferrer"&gt;ComfyUI FLUX.2 template&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/XLabs-AI/x-flux-comfyui" rel="ugc noopener noreferrer"&gt;XLabs ComfyUI extension&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/XLabs-AI/x-flux-comfyui/blob/main/Guide.md" rel="ugc noopener noreferrer"&gt;XLabs node guide&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/city96/ComfyUI-GGUF" rel="ugc noopener noreferrer"&gt;ComfyUI-GGUF&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.bfl.ai/guides/prompting_unified_building" rel="ugc noopener noreferrer"&gt;BFL prompt structure&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.bfl.ai/guides/usecases_t2i_photorealistic" rel="ugc noopener noreferrer"&gt;BFL photorealism guidance&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.bfl.ai/guides/prompting_unified_reference" rel="ugc noopener noreferrer"&gt;BFL camera and lighting reference&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.bfl.ai/guides/prompting_unified_technical" rel="ugc noopener noreferrer"&gt;BFL technical prompting guidance&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>flux</category>
      <category>imagegeneration</category>
      <category>prompts</category>
    </item>
    <item>
      <title>FLUX: the new prodigy text-to-image model</title>
      <dc:creator>Thomas</dc:creator>
      <pubDate>Sat, 03 Aug 2024 07:23:31 +0000</pubDate>
      <link>https://www.promptzone.com/stabletom/flux-the-new-prodigy-model-2ede</link>
      <guid>https://www.promptzone.com/stabletom/flux-the-new-prodigy-model-2ede</guid>
      <description>&lt;p&gt;In a context where generative artificial intelligence is growing exponentially, the text-to-image field never ceases to surprise. After the enthusiastic reception of AuraFlow, it's now Flux's turn to make its grand entrance on the scene of alternatives to Stability AI's &lt;a href="https://www.promptzone.com/deepa_kowalski/ai-image-generators-2026-vheer-visualgpt-fooocus-comfyui-midjourney-more-compared-2i44"&gt;Stable Diffusion&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://blackforestlabs.ai/" rel="nofollow ugc noopener noreferrer"&gt;Black Forest Labs&lt;/a&gt;, a new start-up specializing in generative AI for the media, has just launched Flux, its suite of text-to-image models featuring an open source approach coupled with impressive performance, making it a serious contender not only for Stable Diffusion 3 but also for other industry leaders such as DALL-E and Midjourney.&lt;/p&gt;

&lt;h2 id="flux1"&gt;
  
  
  FLUX.1
&lt;/h2&gt;

&lt;p&gt;Flux's text-image model, trained on a vast dataset of images and captions, has impressive capabilities for generating photorealistic images from natural language descriptions. The model's in-depth understanding of language enables it to interpret complex messages and produce highly detailed, coherent images.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/lfb642s1gpzx10f22ywb.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/lfb642s1gpzx10f22ywb.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2 id="technical-features"&gt;
  
  
  Technical features
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Enhanced image quality:&lt;/strong&gt; Create stunning images at higher resolutions.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Advanced human anatomy and photorealism:&lt;/strong&gt; Get highly realistic, anatomically accurate images.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Faster adhesion:&lt;/strong&gt; Get more accurate and relevant images based on your data.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Exceptional speed:&lt;/strong&gt; Benefit from the speed and efficiency of Flux Schnell, ideal for high-demand applications.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Available in three variants:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;FLUX.1 [dev]: The basic model, shared with a non-commercial license for the community to build on.&lt;/li&gt;
&lt;li&gt;FLUX.1 [schnell]: A distilled version of the basic model, running up to 10 times faster under OpenSource license.&lt;/li&gt;
&lt;li&gt;FLUX.1 [pro]: A private version available only via the API.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/chusxuaiahafq9ry878v.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/chusxuaiahafq9ry878v.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2 id="how-to-use-flux1"&gt;
  
  
  How to use FLUX.1
&lt;/h2&gt;

&lt;h3 id="online-feeds"&gt;
  
  
  Online feeds
&lt;/h3&gt;

&lt;p&gt;If you'd like to try out a few generations with FLUX.1, you can test the different versions on the Fal.AI and Replicate platforms:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Fal.AI&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://fal.ai/models/fal-ai/flux-pro?ref=blog.fal.ai" rel="nofollow ugc noopener noreferrer"&gt;FLUX.1 [pro]&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://fal.ai/models/fal-ai/flux/dev?ref=blog.fal.ai" rel="nofollow ugc noopener noreferrer"&gt;FLUX.1 [dev]&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://fal.ai/models/fal-ai/flux/schnell?ref=blog.fal.ai" rel="nofollow ugc noopener noreferrer"&gt;FLUX.1 [schnell]&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Replicate&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://replicate.com/black-forest-labs/flux-pro" rel="nofollow ugc noopener noreferrer"&gt;FLUX.1 [pro]&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://replicate.com/black-forest-labs/flux-dev" rel="nofollow ugc noopener noreferrer"&gt;FLUX.1 [dev]&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://replicate.com/black-forest-labs/flux-schnell" rel="nofollow ugc noopener noreferrer"&gt;FLUX.1 [schnell]&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The last two models are also shared on HuggingFace where they can be downloaded. &lt;/p&gt;

&lt;p&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-dev/" rel="nofollow ugc noopener noreferrer"&gt;Download FLUX.1 [dev]&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://huggingface.co/black-forest-labs/FLUX.1-schnell/" rel="nofollow ugc noopener noreferrer"&gt;Download FLUX.1 [schnell]&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The Black Forest Labs team has shared a &lt;a href="https://github.com/black-forest-labs/flux" rel="nofollow ugc noopener noreferrer"&gt;sample code on Github&lt;/a&gt; that already allows developers and advanced users to run the models on their own machines.&lt;/p&gt;

&lt;h3 id="flux-in-comfyui"&gt;
  
  
  Flux in ComfyUi
&lt;/h3&gt;

&lt;p&gt;The latest update to &lt;a href="https://www.promptzone.com/jaroslav/how-to-install-and-run-sdxl-models-in-comfyui-a-complete-guide-2nk2"&gt;ComfyUi&lt;/a&gt; should already integrate Flux and enable image generation with [dev] and [schell] models.&lt;/p&gt;

&lt;p&gt;Workflows have been shared and are available on the &lt;a href="https://comfyanonymous.github.io/ComfyUI_examples/flux/" rel="nofollow ugc noopener noreferrer"&gt;Flux example page&lt;/a&gt; on the ComfyUi github.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/l6misd9vh3ytcj4m63s3.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/l6misd9vh3ytcj4m63s3.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2 id="business-strategy-and-licensing"&gt;
  
  
  Business strategy and licensing
&lt;/h2&gt;

&lt;p&gt;Black Forest Labs takes an innovative strategic approach with its FLUX suite, offering options tailored to a variety of needs and usage contexts:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;FLUX.1 [pro]&lt;/strong&gt; : The spearhead of the range, this high-end model promises unrivalled performance. Accessible only via the Black Forest Labs API, it targets professionals demanding exceptional visual quality. Although pricing details have yet to be confirmed, we can expect a premium commercial model.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;FLUX.1 [dev]&lt;/strong&gt; : An intermediate version, FLUX.1 [dev] offers a balance between performance and accessibility. Designed for developers and researchers, it enables fine-tuning and more flexible use, while remaining framed by non-commercial terms of use. For commercial use, specific negotiations with Black Forest Labs will be necessary.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;FLUX.1 [schnell]&lt;/strong&gt;: A true gateway to the FLUX universe, this speed-optimized version sacrifices some quality for accessibility. Distributed under an open source license (Apache 2.0), FLUX.1 [schnell] opens up a vast field of possibilities for large-scale experimentation and innovation.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;This three-pronged strategy illustrates Black Forest Labs' determination to democratize access to generative AI while preserving a viable business model. It thus offers a range of solutions to meet the varied needs of professionals, researchers and AI enthusiasts alike.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/icq8wlhp8z78i96ozsa8.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/icq8wlhp8z78i96ozsa8.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2 id="future-prospects"&gt;
  
  
  &lt;strong&gt;Future prospects&lt;/strong&gt;
&lt;/h2&gt;

&lt;p&gt;The launch of FLUX by Black Forest Labs marks a turning point in the text-to-image industry. Backed by a $31 million Seed round led by Andreessen Horowitz, the company is well positioned to influence the future of generative AI.&lt;/p&gt;

&lt;p&gt;The Black Forest Labs team, made up of renowned researchers and engineers who have contributed to major innovations such as VQGAN and Stable Diffusion, is already planning the development of text-to-video systems. This approach could accelerate innovation in fields as diverse as cinema, advertising and education.&lt;/p&gt;

&lt;p&gt;By democratizing access to these cutting-edge technologies while emphasizing transparency and security, Black Forest Labs aims to shape a more open, collaborative and innovative generative AI ecosystem. Time will tell whether FLUX will be able to hold its own against the giants of the sector, but one thing is certain: competition in the field of text-to-image AI has only just begun.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/faj9jf72997w2fmd42hy.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/faj9jf72997w2fmd42hy.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>aiart</category>
      <category>news</category>
      <category>flux</category>
    </item>
    <item>
      <title>Stable Diffusion Prompt Weights: 2026 Complete Guide</title>
      <dc:creator>Thomas</dc:creator>
      <pubDate>Thu, 01 Aug 2024 13:58:40 +0000</pubDate>
      <link>https://www.promptzone.com/stabletom/varying-prompt-weight-with-stable-diffusion-2nf1</link>
      <guid>https://www.promptzone.com/stabletom/varying-prompt-weight-with-stable-diffusion-2nf1</guid>
      <description>&lt;p&gt;Stable Diffusion prompt weights adjust the influence of selected words: use &lt;code&gt;(word:1.2)&lt;/code&gt; to increase emphasis or &lt;code&gt;(word:0.8)&lt;/code&gt; to reduce it. Plain text has weight 1.0, while &lt;code&gt;(word)&lt;/code&gt; defaults to 1.1 in A1111, ComfyUI, and Fooocus.&lt;/p&gt;

&lt;p&gt;Square brackets, prompt scheduling, and the underlying weighting calculations differ by interface, so use the syntax table before copying a prompt between tools. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Sources: A1111 parser&lt;/a&gt;, &lt;a href="https://docs.comfy.org/built-in-nodes/ClipTextEncode" rel="ugc noopener noreferrer"&gt;ComfyUI documentation&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Fooocus parser&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="what-is-prompt-weighting"&gt;
  
  
  What is Prompt Weighting?
&lt;/h2&gt;

&lt;p&gt;Prompt weighting changes how parts of your description contribute to the text conditioning used for generation. The interface parses your syntax before the model receives that conditioning. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/sd_emphasis.py" rel="ugc noopener noreferrer"&gt;Sources: A1111 emphasis&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;ComfyUI encoder&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;In a prompt such as &lt;code&gt;woman, beach, pizza&lt;/code&gt;, unweighted text receives a multiplier of 1.0. This does &lt;strong&gt;not&lt;/strong&gt; mean each word owns one-third of the picture or one-third of the model's attention. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Sources: parsers&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;ComfyUI encoder&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The weights are controls on encoded text, not percentages of objects or guaranteed screen space. A larger coefficient does not specify the size, location, or count of the named object.&lt;/p&gt;

&lt;p&gt;If you are choosing an interface, our &lt;a href="https://www.promptzone.com/deepa_kowalski/ai-image-generators-2026-vheer-visualgpt-fooocus-comfyui-midjourney-more-compared-2i44"&gt;AI image generator comparison&lt;/a&gt; provides broader context.&lt;/p&gt;

&lt;h3 id="increasing-word-weights"&gt;
  
  
  Increasing word weights
&lt;/h3&gt;

&lt;p&gt;Start with an unweighted description and change one phrase:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;woman eating pizza on a beach
woman eating pizza on a (beach:1.2)
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The second prompt assigns the beach phrase a higher weight. Compare framing, background, and subject placement; do not judge the result solely by whether a beach appears. The syntax follows the &lt;a href="https://docs.comfy.org/built-in-nodes/ClipTextEncode" rel="ugc noopener noreferrer"&gt;CLIP Text Encode documentation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/xwvkx2p459ru3a8pis2n.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/xwvkx2p459ru3a8pis2n.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Compare the illustrated compositions, then repeat the experiment in your interface with the model and generation settings held constant.&lt;/p&gt;

&lt;h3 id="every-word-counts"&gt;
  
  
  Every word counts
&lt;/h3&gt;

&lt;p&gt;A longer prompt does not divide attention into equal shares. A word can span multiple tokens, and its encoded representation includes context from surrounding text. &lt;a href="https://github.com/BlenderNeko/ComfyUI_ADV_CLIP_emb" rel="ugc noopener noreferrer"&gt;Source: Advanced CLIP documentation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Try a more specific description:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;photograph of a woman with dark hair eating pepperoni pizza
on a (white beach:1.2), sunset, ocean, natural light
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Here the modifier covers &lt;code&gt;white beach&lt;/code&gt; together. Keep the rest of the description unchanged during the comparison so that you can assess the contribution of that phrase.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/r43tgzqxtb5bcnby7h3m.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/r43tgzqxtb5bcnby7h3m.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;If the composition becomes less useful, simplify the wording or remove competing details before increasing the coefficient again.&lt;/p&gt;

&lt;h2 id="syntax-by-ui"&gt;
  
  
  Syntax by UI
&lt;/h2&gt;

&lt;p&gt;This table describes standard A1111 emphasis and the built-in Stable Diffusion CLIP text paths in ComfyUI and Fooocus. Custom nodes or extensions can introduce different parsers.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Interface&lt;/th&gt;
&lt;th&gt;Increase emphasis&lt;/th&gt;
&lt;th&gt;Reduce emphasis&lt;/th&gt;
&lt;th&gt;Square brackets&lt;/th&gt;
&lt;th&gt;
&lt;code&gt;AND&lt;/code&gt; / &lt;code&gt;[a:b:0.5]&lt;/code&gt;
&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;AUTOMATIC1111&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;(word)&lt;/code&gt; = 1.1; &lt;code&gt;(word:1.3)&lt;/code&gt; is explicit&lt;/td&gt;
&lt;td&gt;&lt;code&gt;(word:0.8)&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;[word]&lt;/code&gt; = 1/1.1&lt;/td&gt;
&lt;td&gt;Supported as separate composition / scheduling features&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;ComfyUI built-in CLIP&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;(word)&lt;/code&gt; = 1.1; &lt;code&gt;(word:1.2)&lt;/code&gt; is explicit&lt;/td&gt;
&lt;td&gt;&lt;code&gt;(word:0.8)&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;No decrement; treated as text&lt;/td&gt;
&lt;td&gt;Not interpreted as those operators by the built-in text encoder&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Fooocus&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;(word)&lt;/code&gt; = 1.1; &lt;code&gt;(word:1.2)&lt;/code&gt; is explicit&lt;/td&gt;
&lt;td&gt;&lt;code&gt;(word:0.8)&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;No decrement in its parser&lt;/td&gt;
&lt;td&gt;Not supported as A1111 composition / scheduling operators in the standard path&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;A1111's advanced operators are documented in its &lt;a href="https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features" rel="ugc noopener noreferrer"&gt;Features wiki&lt;/a&gt;. Fooocus's &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/modules/async_worker.py" rel="ugc noopener noreferrer"&gt;generation worker&lt;/a&gt; processes prompts without A1111's composition or step-scheduling parser.&lt;/p&gt;

&lt;h2 id="how-do-i-use-it"&gt;
  
  
  How do I use it?
&lt;/h2&gt;

&lt;p&gt;Choose the smallest phrase that describes your intended change. Generate a baseline, apply one explicit weight, and compare while keeping the checkpoint, seed, dimensions, sampler, steps, and other prompt text unchanged.&lt;/p&gt;

&lt;p&gt;Try another seed before adopting the change. Evaluate each coefficient against your intended result.&lt;/p&gt;

&lt;h3 id="prompt-weight-in-automatic1111"&gt;
  
  
  Prompt Weight in Automatic1111
&lt;/h3&gt;

&lt;p&gt;A1111 supports both emphasis shortcuts and explicit coefficients. Its default &lt;strong&gt;Emphasis mode&lt;/strong&gt; is &lt;strong&gt;Original&lt;/strong&gt;; other modes can change or disable weighting behavior. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/shared_options.py" rel="ugc noopener noreferrer"&gt;Source: settings definitions&lt;/a&gt;.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;(blue eyes)&lt;/code&gt; applies a 1.1 multiplier.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;((blue eyes))&lt;/code&gt; applies 1.1 × 1.1, or 1.21.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;(blue eyes:1.3)&lt;/code&gt; supplies the coefficient directly.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;[blue eyes]&lt;/code&gt; applies 1/1.1, approximately 0.9091.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;(blue eyes:0.8)&lt;/code&gt; explicitly reduces the weight to 0.8.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;These values follow the &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;A1111 attention parser&lt;/a&gt;. Square-bracket reduction is not exactly 0.9. Use an explicit numeric weight when that distinction matters to your comparison.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/kf4r5vxvnpnme9juryik.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/kf4r5vxvnpnme9juryik.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Numeric weights belong inside parentheses. &lt;code&gt;[blue eyes:0.8]&lt;/code&gt; is not the documented numeric emphasis form and can be interpreted as prompt editing. &lt;a href="https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features" rel="ugc noopener noreferrer"&gt;Source: A1111 Features&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For literal parentheses, escape them:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;portrait of a character \(adult\), blue jacket
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The backslashes preserve the parentheses as text instead of starting an emphasis group. A1111 also accepts escaped square brackets, &lt;code&gt;\[&lt;/code&gt; and &lt;code&gt;\]&lt;/code&gt;. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Source: A1111 parser&lt;/a&gt;.&lt;/p&gt;

&lt;h4 id="and-combines-prompt-conditionings"&gt;
  
  
  AND combines prompt conditionings
&lt;/h4&gt;

&lt;p&gt;Uppercase &lt;code&gt;AND&lt;/code&gt; is A1111's composable-diffusion operator. It separates prompts, and each part can have its own coefficient:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;a lighthouse:1.2 AND a stormy coastline:0.8
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;This combines separately conditioned prompts; it is not equivalent to emphasizing one phrase in a single prompt. It also does not assign the lighthouse to a particular region of the canvas. &lt;a href="https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features" rel="ugc noopener noreferrer"&gt;Source: composable diffusion documentation&lt;/a&gt;.&lt;/p&gt;

&lt;h4 id="prompt-editing-changes-text-during-sampling"&gt;
  
  
  Prompt editing changes text during sampling
&lt;/h4&gt;

&lt;p&gt;For A1111's initial sampling pass:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;a [red:blue:0.5] ceramic bowl on a table
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The prompt uses &lt;code&gt;red&lt;/code&gt; for the first half of the steps and &lt;code&gt;blue&lt;/code&gt; afterward. The value 0.5 specifies a switching point, not a weight or an equal color mixture. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Source: scheduling implementation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Integer timing values specify steps; decimals specify fractions. Hires-fix scheduling has a separate timeline, so check its rules before transferring a first-pass schedule. &lt;a href="https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features" rel="ugc noopener noreferrer"&gt;Source: A1111 prompt editing&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="prompt-weight-in-fooocus"&gt;
  
  
  Prompt Weight in Fooocus
&lt;/h3&gt;

&lt;p&gt;&lt;a href="https://www.promptzone.com/jaroslav/how-to-use-fooocus-a-practical-guide-and-tricks-3hfk"&gt;Fooocus&lt;/a&gt; supports &lt;code&gt;(phrase:weight)&lt;/code&gt;, plain parentheses, and escaped literal parentheses. Its README explicitly documents weighted phrases. &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Sources: README&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;parser&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For a reduced weight, use &lt;code&gt;(phrase:0.8)&lt;/code&gt;. Its parser does not implement A1111's square-bracket decrement, even though the README describes its reweighting as A1111-compatible. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Sources: parser&lt;/a&gt;, &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;README&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Select text and use &lt;strong&gt;Ctrl+Up&lt;/strong&gt; or &lt;strong&gt;Ctrl+Down&lt;/strong&gt; to edit attention. Inspect the resulting text before generating. &lt;a href="https://github.com/lllyasviel/Fooocus/blob/main/update_log.md#21776" rel="ugc noopener noreferrer"&gt;Source: change log&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Keep styles fixed during a weighting comparison. Fooocus can add style text and prompt expansion, so a changed style selection introduces another difference alongside your coefficient. &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Sources: README&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/modules/async_worker.py" rel="ugc noopener noreferrer"&gt;generation worker&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="prompt-weight-in-comfyui"&gt;
  
  
  Prompt Weight in ComfyUI
&lt;/h3&gt;

&lt;p&gt;In &lt;a href="https://www.promptzone.com/jaroslav/how-to-install-and-run-sdxl-models-in-comfyui-a-complete-guide-2nk2"&gt;ComfyUI&lt;/a&gt;, enter &lt;code&gt;(phrase:1.2)&lt;/code&gt; in the CLIP Text Encode node for increased emphasis, or &lt;code&gt;(phrase:0.8)&lt;/code&gt; for reduced emphasis. Plain &lt;code&gt;(phrase)&lt;/code&gt; means 1.1. &lt;a href="https://docs.comfy.org/built-in-nodes/ClipTextEncode" rel="ugc noopener noreferrer"&gt;Source: node documentation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The built-in parser does not use &lt;code&gt;[phrase]&lt;/code&gt; for reduction. It also does not turn A1111's &lt;code&gt;AND&lt;/code&gt; or &lt;code&gt;[a:b:0.5]&lt;/code&gt; into composition or scheduling operations. Those require an appropriate workflow or extension. &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Source: encoder implementation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;You can select prompt text and press &lt;strong&gt;Ctrl+Up/Down&lt;/strong&gt; to adjust its weight. The adjustment size is configurable. Escape literal parentheses with &lt;code&gt;\(&lt;/code&gt; and &lt;code&gt;\)&lt;/code&gt;. &lt;a href="https://docs.comfy.org/built-in-nodes/ClipTextEncode" rel="ugc noopener noreferrer"&gt;Source: node documentation&lt;/a&gt;.&lt;/p&gt;

&lt;h4 id="advanced-clip-encode-weightinterpretation"&gt;
  
  
  Advanced CLIP Encode weight_interpretation
&lt;/h4&gt;

&lt;p&gt;BlenderNeko's &lt;strong&gt;CLIP Text Encode (Advanced)&lt;/strong&gt; is a custom node, separate from the built-in encoder. Its &lt;code&gt;weight_interpretation&lt;/code&gt; options are &lt;code&gt;comfy&lt;/code&gt;, &lt;code&gt;A1111&lt;/code&gt;, &lt;code&gt;compel&lt;/code&gt;, &lt;code&gt;comfy++&lt;/code&gt;, and &lt;code&gt;down_weight&lt;/code&gt;. &lt;a href="https://github.com/BlenderNeko/ComfyUI_ADV_CLIP_emb" rel="ugc noopener noreferrer"&gt;Source: extension README&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Its separate &lt;code&gt;token_normalization&lt;/code&gt; setting offers &lt;code&gt;none&lt;/code&gt;, &lt;code&gt;mean&lt;/code&gt;, &lt;code&gt;length&lt;/code&gt;, and &lt;code&gt;length+mean&lt;/code&gt;. Record both settings when sharing a workflow; selecting an interpretation changes how the node handles the weights. &lt;a href="https://github.com/BlenderNeko/ComfyUI_ADV_CLIP_emb" rel="ugc noopener noreferrer"&gt;Source: extension README&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="difference-between-automatic1111-and-comfyui"&gt;
  
  
  Difference between Automatic1111 and ComfyUI
&lt;/h2&gt;

&lt;p&gt;A1111's Original emphasis multiplies encoded token vectors by their coefficients, then rescales the result to restore the original mean. It does not normalize word weights so they sum to one. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/sd_emphasis.py" rel="ugc noopener noreferrer"&gt;Source: emphasis implementation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;ComfyUI's built-in encoder instead applies weighting relative to an empty-prompt embedding. In simplified form, it computes &lt;code&gt;empty + weight × (token − empty)&lt;/code&gt; for a weighted token. &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Source: ComfyUI encoder&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Fooocus uses that empty-prompt-relative operation, then restores the original mean. Its README's compatibility wording does not establish identical computations or images. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/modules/patch_clip.py" rel="ugc noopener noreferrer"&gt;Source: Fooocus CLIP patch&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;These implementation differences mean shared syntax does not establish equivalent encodings. Treat a copied weight as a starting point to evaluate in the destination workflow.&lt;/p&gt;

&lt;h2 id="worked-examples"&gt;
  
  
  Worked examples
&lt;/h2&gt;

&lt;h3 id="emphasize-a-complete-attribute"&gt;
  
  
  Emphasize a complete attribute
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;portrait of a woman wearing a (blue linen jacket:1.2), daylight
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Compare with the unweighted prompt. Inspect whether the jacket, color, and overall composition remain useful.&lt;/p&gt;

&lt;h3 id="reduce-emphasis-without-removing-the-concept"&gt;
  
  
  Reduce emphasis without removing the concept
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;mountain landscape, (mist:0.8), pine forest, morning light
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Compare the reduced-weight version with &lt;code&gt;mist&lt;/code&gt; unweighted. Down-weighting changes conditioning; it does not remove the word from the description. If you do not want mist at all, first try deleting that phrase. &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Source: encoder&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="separate-timing-from-strength"&gt;
  
  
  Separate timing from strength
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;studio photograph of a [glass:ceramic:0.5] vase
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Use this only with A1111's scheduling support. It asks for an early glass-conditioned phase followed by a ceramic-conditioned phase; it does not assign either material a 0.5 emphasis weight. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Source: scheduler&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="common-mistakes"&gt;
  
  
  Common mistakes
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Copying square brackets into ComfyUI or Fooocus:&lt;/strong&gt; use explicit &lt;code&gt;(phrase:0.8)&lt;/code&gt; reduction for the documented built-in paths.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Weighting everything:&lt;/strong&gt; choose the phrase you want to investigate so that the comparison stays interpretable.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Changing several controls together:&lt;/strong&gt; hold other settings fixed while evaluating a coefficient.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Expecting object placement or percentages:&lt;/strong&gt; token weights do not define layout regions or image-area shares.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Assuming nested numeric groups transfer exactly:&lt;/strong&gt; use flat explicit groups when moving prompts between different parsers.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ignoring the active encoder:&lt;/strong&gt; custom-node interpretation and normalization settings can change the result.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The syntax cautions follow the &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;A1111&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;ComfyUI&lt;/a&gt;, and &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Fooocus&lt;/a&gt; parsers. The &lt;a href="https://github.com/BlenderNeko/ComfyUI_ADV_CLIP_emb" rel="ugc noopener noreferrer"&gt;Advanced CLIP README&lt;/a&gt; explains why encoder settings matter.&lt;/p&gt;

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

&lt;h3 id="what-is-the-default-stable-diffusion-prompt-weight"&gt;
  
  
  What is the default Stable Diffusion prompt weight?
&lt;/h3&gt;

&lt;p&gt;Unweighted text uses 1.0 in these parsers. Plain parentheses apply 1.1; an explicit coefficient such as &lt;code&gt;(word:1.2)&lt;/code&gt; selects that value. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Sources: A1111&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;ComfyUI&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Fooocus&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="do-brackets-decrease-prompt-weight"&gt;
  
  
  Do brackets decrease prompt weight?
&lt;/h3&gt;

&lt;p&gt;In A1111, &lt;code&gt;[word]&lt;/code&gt; multiplies by 1/1.1. ComfyUI and Fooocus do not implement that decrement. Use &lt;code&gt;(word:0.8)&lt;/code&gt; for explicit reduction across all three. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Sources: A1111&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;ComfyUI&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Fooocus&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="what-prompt-weight-gives-the-best-results"&gt;
  
  
  What prompt weight gives the best results?
&lt;/h3&gt;

&lt;p&gt;No value is universally best. Compare an unweighted phrase with a modest explicit change, and choose according to your intended image. Keep the interface, encoder configuration, and other generation settings consistent during the test.&lt;/p&gt;

&lt;h3 id="is-raw-ab05-endraw-the-same-as-raw-b05-endraw-"&gt;
  
  
  Is &lt;code&gt;[a:b:0.5]&lt;/code&gt; the same as &lt;code&gt;(b:0.5)&lt;/code&gt;?
&lt;/h3&gt;

&lt;p&gt;No. In A1111, the first schedules a change from &lt;code&gt;a&lt;/code&gt; to &lt;code&gt;b&lt;/code&gt; halfway through the initial sampling pass. The second reduces the weight of &lt;code&gt;b&lt;/code&gt; throughout its use in the prompt. &lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;Source: A1111 parser&lt;/a&gt;.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features" rel="ugc noopener noreferrer"&gt;AUTOMATIC1111 Features: emphasis, composition, prompt editing&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/prompt_parser.py" rel="ugc noopener noreferrer"&gt;AUTOMATIC1111 attention and scheduling parser&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/sd_emphasis.py" rel="ugc noopener noreferrer"&gt;AUTOMATIC1111 emphasis calculations&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/modules/shared_options.py" rel="ugc noopener noreferrer"&gt;AUTOMATIC1111 default emphasis setting&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.comfy.org/built-in-nodes/ClipTextEncode" rel="ugc noopener noreferrer"&gt;ComfyUI CLIP Text Encode documentation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/Comfy-Org/ComfyUI/master/comfy/sd1_clip.py" rel="ugc noopener noreferrer"&gt;ComfyUI parser and token-weight encoder&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/BlenderNeko/ComfyUI_ADV_CLIP_emb" rel="ugc noopener noreferrer"&gt;Advanced CLIP Text Encode extension documentation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Fooocus README&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/ldm_patched/modules/sd1_clip.py" rel="ugc noopener noreferrer"&gt;Fooocus prompt parser&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/modules/patch_clip.py" rel="ugc noopener noreferrer"&gt;Fooocus CLIP weighting implementation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/blob/main/update_log.md#21776" rel="ugc noopener noreferrer"&gt;Fooocus attention keyboard shortcuts&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/modules/async_worker.py" rel="ugc noopener noreferrer"&gt;Fooocus generation and prompt-processing path&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;

&lt;h2 id="free-tools-for-this"&gt;
  
  
  Free tools for this
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.promptzone.com/prompt-weight-converter"&gt;Stable Diffusion Prompt Weight Converter&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.promptzone.com/negative-prompt-generator"&gt;Negative Prompt Generator for Stable Diffusion&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>ai</category>
      <category>stablediffusion</category>
      <category>prompts</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>Fooocus 2.5: What Changed, New Features, and How to Update</title>
      <dc:creator>Thomas</dc:creator>
      <pubDate>Tue, 30 Jul 2024 18:49:38 +0000</pubDate>
      <link>https://www.promptzone.com/stabletom/fooocus-25-is-here-16p4</link>
      <guid>https://www.promptzone.com/stabletom/fooocus-25-is-here-16p4</guid>
      <description>&lt;p&gt;Fooocus 2.5 introduced automatic Enhance workflows, improved masking, and a Pony preset. The latest official tagged release is &lt;strong&gt;v2.5.5, published August 12, 2024&lt;/strong&gt;; the project is in limited long-term support with bug fixes only.&lt;/p&gt;

&lt;p&gt;Update through the official launcher, or fetch the code and install its requirements manually if automatic updating fails. &lt;a href="https://github.com/lllyasviel/Fooocus/releases" rel="ugc noopener noreferrer"&gt;Sources: releases&lt;/a&gt;, &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;project status&lt;/a&gt;, &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;update instructions&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="fooocus-latest-version-and-maintenance-status"&gt;
  
  
  Fooocus latest version and maintenance status
&lt;/h2&gt;

&lt;p&gt;Fooocus 2.5.0 was published on July 17, 2024; v2.5.5 remains the latest tagged release. &lt;a href="https://github.com/lllyasviel/Fooocus/releases" rel="ugc noopener noreferrer"&gt;Source: releases&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Fooocus is an SDXL project under limited support, with no current plans for newer model architectures. A model's availability elsewhere does not establish Fooocus support. &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Source: README&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The 2.5 release notes credit substantial work by &lt;a href="https://github.com/mashb1t" rel="ugc noopener noreferrer"&gt;mashb1t&lt;/a&gt;. For everyday operation, our &lt;a href="https://www.promptzone.com/jaroslav/how-to-use-fooocus-a-practical-guide-and-tricks-3hfk"&gt;Fooocus practical guide&lt;/a&gt; covers the interface. &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;Source: v2.5.0&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;To compare other tools, see the &lt;a href="https://www.promptzone.com/deepa_kowalski/ai-image-generators-2026-vheer-visualgpt-fooocus-comfyui-midjourney-more-compared-2i44"&gt;AI image generator overview&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="what-is-new-in-25"&gt;
  
  
  What is new in 2.5
&lt;/h2&gt;

&lt;p&gt;The initial v2.5.0 release added:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Enhance:&lt;/strong&gt; automatic image refinement using detection prompts and replacement prompts.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Masking improvements:&lt;/strong&gt; GroundingDINO/SAM changes and an editor for uploaded masks.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Model and preset changes:&lt;/strong&gt; AnimaPencilXL v500, &lt;code&gt;DreamShaperXL_Turbo_v2_1&lt;/code&gt;, Film Photography Style V1, plus the Pony preset and style.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Configuration improvements:&lt;/strong&gt; preset VAE support and a configurable default inpaint engine.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Sampling and metadata changes:&lt;/strong&gt; a restart sampler and persistent model metadata hash caching.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;These changes are listed in the &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;v2.5.0 release notes&lt;/a&gt;. Later 2.5.x releases added smaller improvements and fixes:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Release&lt;/th&gt;
&lt;th&gt;Notable change&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.1" rel="ugc noopener noreferrer"&gt;2.5.1&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Restored JPEG/JPG/WebP metadata reading; added saving only the final enhanced image and improved Enhance image ordering&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.2" rel="ugc noopener noreferrer"&gt;2.5.2&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Fixed positive prompts with certain styles; expanded input-image configuration&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.3" rel="ugc noopener noreferrer"&gt;2.5.3&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Changed loading to use &lt;code&gt;weights_only&lt;/code&gt;; added control over style changes when describing images&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.4" rel="ugc noopener noreferrer"&gt;2.5.4&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Fixed configuration validation, directory searching, and Enhance debug-mask previews&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.5" rel="ugc noopener noreferrer"&gt;2.5.5&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;Fixed an unsupported image-type problem in Colab&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2 id="enhance-correct-and-improve"&gt;
  
  
  Enhance: Correct and Improve
&lt;/h2&gt;

&lt;p&gt;Enhance can upscale an image and refine detected regions automatically. Its detection prompt identifies the area to work on; separate positive and negative prompts describe the intended replacement. &lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Source: Enhance documentation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For example, try detecting a dress and describing the desired garment color. Treat the output as a candidate to inspect: check the mask and surrounding details before accepting the edit.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/1i8a5ylv7x8i3tzxri6b.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/1i8a5ylv7x8i3tzxri6b.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;For a generated image, keep the seed fixed, open Enhance, and enable the refinement steps you need. For an existing image, open &lt;strong&gt;Input Image → Enhance&lt;/strong&gt; and upload it. &lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Source: Enhance documentation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://promptzone-community.s3.amazonaws.com/uploads/articles/9z9jlky367v29phbvrx6.png" class="article-body-image-wrapper"&gt;&lt;img src="https://promptzone-community.s3.amazonaws.com/uploads/articles/9z9jlky367v29phbvrx6.png" alt=" "&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The default masking combination uses GroundingDINO to locate objects and SAM to segment them. Additional masking options are available. This separates finding the target from describing what should replace it. &lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Source: Enhance documentation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Upscaling can run before or after enhancement. The maintainer recommends the earlier position for small details, while enhancing first can suit larger regions such as clothing or backgrounds. &lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Source: Enhance documentation&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="new-models-and-presets"&gt;
  
  
  New Models and Presets
&lt;/h2&gt;

&lt;p&gt;The Anime preset uses &lt;a href="https://civitai.com/models/261336/animapencil-xl" rel="ugc noopener noreferrer"&gt;AnimaPencilXL v500&lt;/a&gt;. Its current JSON selects &lt;code&gt;animaPencilXL_v500.safetensors&lt;/code&gt;; an older model already present can be used through the preset's fallback list. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/presets/anime.json" rel="ugc noopener noreferrer"&gt;Source: anime configuration&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The Realistic preset uses &lt;code&gt;realisticStockPhoto_v20.safetensors&lt;/code&gt; with &lt;a href="https://civitai.com/models/158945/sdxl-film-photography-style" rel="ugc noopener noreferrer"&gt;SDXL Film Photography Style V1&lt;/a&gt; at weight 0.25. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/presets/realistic.json" rel="ugc noopener noreferrer"&gt;Source: realistic configuration&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The added &lt;code&gt;pony_v6&lt;/code&gt; preset selects &lt;a href="https://civitai.com/models/257749/pony-diffusion-v6-xl" rel="ugc noopener noreferrer"&gt;Pony Diffusion V6 XL&lt;/a&gt;, its VAE, and the Fooocus Pony style. The preset sets the inpaint engine to &lt;code&gt;None&lt;/code&gt;. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/presets/pony_v6.json" rel="ugc noopener noreferrer"&gt;Source: Pony configuration&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Fooocus Pony adds &lt;code&gt;score_9&lt;/code&gt;, &lt;code&gt;score_8_up&lt;/code&gt;, and &lt;code&gt;score_7_up&lt;/code&gt; to the positive prompt. Its negative style contains &lt;code&gt;score_6&lt;/code&gt;, &lt;code&gt;score_5&lt;/code&gt;, and &lt;code&gt;score_4&lt;/code&gt;; those tags are supplied by the style itself. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/sdxl_styles/sdxl_styles_fooocus.json" rel="ugc noopener noreferrer"&gt;Source: style definitions&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Playground v2.5 support and its preset arrived in v2.4.2, with further preset adjustments in v2.5.0. This distinction matters when checking which feature requires which version. &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.4.2" rel="ugc noopener noreferrer"&gt;Sources: v2.4.2&lt;/a&gt;, &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;v2.5.0&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="other-notable-updates"&gt;
  
  
  Other Notable Updates
&lt;/h2&gt;

&lt;p&gt;The inpainting tab contains &lt;strong&gt;Enable Advanced Masking Features&lt;/strong&gt;, the renamed mask-upload control. The mask inversion control also moved out of Developer Debug Mode, and uploaded masks gained editing functionality. &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;Source: v2.5.0&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Presets can specify a VAE and default inpaint engine. Inspect the supplied configuration when changing models: checkpoints may require different inpainting settings. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/modules/config.py" rel="ugc noopener noreferrer"&gt;Sources: preset keys&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/presets/pony_v6.json" rel="ugc noopener noreferrer"&gt;Pony&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="how-to-update"&gt;
  
  
  How to update
&lt;/h2&gt;

&lt;h3 id="update-fooocus-through-its-launcher"&gt;
  
  
  Update Fooocus through its launcher
&lt;/h3&gt;

&lt;p&gt;Close Fooocus and back up custom presets, configuration, and any edited project files. For the official Windows package, launch &lt;code&gt;run.bat&lt;/code&gt;; its generated command invokes &lt;code&gt;Fooocus\entry_with_update.py&lt;/code&gt;. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/build_launcher.py" rel="ugc noopener noreferrer"&gt;Source: launcher implementation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For a Python installation, activate the environment you use for Fooocus, open its directory, and 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 entry_with_update.py
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The updater fetches the current branch from &lt;code&gt;origin&lt;/code&gt; and attempts a fast-forward update. Startup then checks dependencies against &lt;code&gt;requirements_versions.txt&lt;/code&gt;. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/entry_with_update.py" rel="ugc noopener noreferrer"&gt;Sources: updater&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/launch.py" rel="ugc noopener noreferrer"&gt;startup dependency handling&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="update-manually-if-automatic-updating-fails"&gt;
  
  
  Update manually if automatic updating fails
&lt;/h3&gt;

&lt;p&gt;From the Fooocus repository directory, with its Python environment active:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git pull
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_versions.txt
python entry_with_update.py
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;git pull&lt;/code&gt; fetches and integrates code changes; the pip command installs the required packages. Installing requirements alone does not update Fooocus's source code. &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;Sources: release update procedure&lt;/a&gt;, &lt;a href="https://git-scm.com/docs/git-pull" rel="ugc noopener noreferrer"&gt;Git documentation&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;For the Windows archive installation, open Command Prompt in the inner &lt;code&gt;Fooocus&lt;/code&gt; folder containing &lt;code&gt;config.txt&lt;/code&gt;. The release's dependency command uses the bundled interpreter:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight batchfile"&gt;&lt;code&gt;..\python_embeded\python.exe &lt;span class="na"&gt;-m &lt;/span&gt;&lt;span class="kd"&gt;pip&lt;/span&gt; &lt;span class="kd"&gt;install&lt;/span&gt; &lt;span class="na"&gt;-r &lt;/span&gt;.\requirements_versions.txt
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Run &lt;code&gt;git pull&lt;/code&gt; there first if Git is installed, then restart through &lt;code&gt;run.bat&lt;/code&gt;. If you cannot repair that installation, the release notes also offer downloading the official package again. &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;Source: v2.5.0 update instructions&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="windows-package-names-and-runupdate"&gt;
  
  
  Windows package names and run_update
&lt;/h3&gt;

&lt;p&gt;The README links the Windows archive &lt;code&gt;Fooocus_win64_2-5-0.7z&lt;/code&gt;. That archive name differs from the latest release tag because the launcher handles code updates. &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Sources: Windows download&lt;/a&gt;, &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/entry_with_update.py" rel="ugc noopener noreferrer"&gt;updater&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The official launcher generator creates &lt;code&gt;run.bat&lt;/code&gt;, &lt;code&gt;run_anime.bat&lt;/code&gt;, and &lt;code&gt;run_realistic.bat&lt;/code&gt;; it does not create &lt;code&gt;run_update&lt;/code&gt;. Use the documented launcher or &lt;code&gt;entry_with_update.py&lt;/code&gt; for the official package. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/build_launcher.py" rel="ugc noopener noreferrer"&gt;Source: launcher generator&lt;/a&gt;.&lt;/p&gt;

&lt;h2 id="known-issues"&gt;
  
  
  Known issues
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Update failure:&lt;/strong&gt; local changes or a diverged branch can prevent a fast-forward update. Inspect the terminal error and preserve your edits before resolving the checkout. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/entry_with_update.py" rel="ugc noopener noreferrer"&gt;Source: updater&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Misleading success text:&lt;/strong&gt; the updater prints “Update succeeded.” even after its exception handler. Read preceding messages when confirming that an update completed. &lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/entry_with_update.py" rel="ugc noopener noreferrer"&gt;Source: updater&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Enhance compatibility:&lt;/strong&gt; the maintainer recommends Speed or Quality for workflows using the inpaint engine; performance LoRAs can interact poorly with it. &lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Source: Enhance documentation&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Control inputs:&lt;/strong&gt; ImagePrompt, PyraCanny, CPDS, and FaceSwap are documented as unsupported within Enhance steps, though they can help generate the starting image. &lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Source: Enhance documentation&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;For unexpected edits, inspect the detection mask first. Model-load errors &lt;code&gt;MetadataIncompleteBuffer&lt;/code&gt; and &lt;code&gt;PytorchStreamReader&lt;/code&gt; indicate corrupted model downloads. &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Source: README&lt;/a&gt;.&lt;/p&gt;

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

&lt;h3 id="what-is-the-fooocus-latest-version"&gt;
  
  
  What is the Fooocus latest version?
&lt;/h3&gt;

&lt;p&gt;The latest official tagged release is v2.5.5, published August 12, 2024. Consult the official release list when checking a different installation or package. &lt;a href="https://github.com/lllyasviel/Fooocus/releases" rel="ugc noopener noreferrer"&gt;Source: releases&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="is-fooocus-still-maintained"&gt;
  
  
  Is Fooocus still maintained?
&lt;/h3&gt;

&lt;p&gt;Its stated status is limited long-term support focused on bug fixes. The README does not announce a migration to newer model architectures. &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Source: project status&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="does-fooocus-25-add-flux-support"&gt;
  
  
  Does Fooocus 2.5 add FLUX support?
&lt;/h3&gt;

&lt;p&gt;No. The official project remains based on SDXL, and its README directs people interested in FLUX to other interfaces. &lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Source: README&lt;/a&gt;.&lt;/p&gt;

&lt;h3 id="do-i-need-a-separate-extension-for-enhance"&gt;
  
  
  Do I need a separate extension for Enhance?
&lt;/h3&gt;

&lt;p&gt;No. Enhance is part of Fooocus 2.5.0 and later. Its built-in workflow accepts generated or uploaded images and uses configurable refinement steps. &lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;Sources: release&lt;/a&gt;, &lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Enhance documentation&lt;/a&gt;.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases" rel="ugc noopener noreferrer"&gt;Official release history&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus" rel="ugc noopener noreferrer"&gt;Fooocus README and maintenance status&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.0" rel="ugc noopener noreferrer"&gt;v2.5.0 features and update procedure&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.1" rel="ugc noopener noreferrer"&gt;v2.5.1 release&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.2" rel="ugc noopener noreferrer"&gt;v2.5.2 release&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.3" rel="ugc noopener noreferrer"&gt;v2.5.3 release&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.4" rel="ugc noopener noreferrer"&gt;v2.5.4 release&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.5.5" rel="ugc noopener noreferrer"&gt;v2.5.5 release&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/releases/tag/v2.4.2" rel="ugc noopener noreferrer"&gt;v2.4.2 release&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/lllyasviel/Fooocus/discussions/3281" rel="ugc noopener noreferrer"&gt;Maintainer's Enhance documentation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/presets/anime.json" rel="ugc noopener noreferrer"&gt;Anime preset&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/presets/realistic.json" rel="ugc noopener noreferrer"&gt;Realistic preset&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/presets/pony_v6.json" rel="ugc noopener noreferrer"&gt;Pony preset&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/sdxl_styles/sdxl_styles_fooocus.json" rel="ugc noopener noreferrer"&gt;Fooocus style definitions&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/modules/config.py" rel="ugc noopener noreferrer"&gt;Preset configuration keys&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/build_launcher.py" rel="ugc noopener noreferrer"&gt;Official Windows launcher generator&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/entry_with_update.py" rel="ugc noopener noreferrer"&gt;Automatic updater implementation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://raw.githubusercontent.com/lllyasviel/Fooocus/main/launch.py" rel="ugc noopener noreferrer"&gt;Startup dependency handling&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://git-scm.com/docs/git-pull" rel="ugc noopener noreferrer"&gt;Git pull documentation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/mashb1t" rel="ugc noopener noreferrer"&gt;Contributor profile&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://civitai.com/models/261336/animapencil-xl" rel="ugc noopener noreferrer"&gt;AnimaPencil model page&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://civitai.com/models/158945/sdxl-film-photography-style" rel="ugc noopener noreferrer"&gt;Film Photography Style model page&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://civitai.com/models/257749/pony-diffusion-v6-xl" rel="ugc noopener noreferrer"&gt;Pony Diffusion model page&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>fooocus</category>
      <category>stablediffusion</category>
      <category>news</category>
    </item>
  </channel>
</rss>
