Note: the last version of the workflow has been greatly simplified to use the ComfyUI sub-workflow system. In consequence, it has less features than the previous ones, but is more stable. I will try to re-include these features over time.<h2 id="introduction">Introduction</h2>Here's my Scene Composer worklfow for <a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI">ComfyUI</a>.The main goal is to create short 5-panels stories in just one queue. For that, it chose randomly parts of the prompt that are used for generation, like:<ul><li>Character (e.g. hair, eyes, attitude)</li><li>Clothes &amp; Underwear</li><li>Sexual position and action</li></ul>To keep consistency, it also keeps certain parts of the prompt to inject it across all scenes (like the environment, the main character and their clothes). You can read my <a target="_blank" rel="ugc" href="https://civarchive.com/articles/5760/scene-composer-overview-and-usecases">Overview &amp; Usecases</a> article for more explanations!If you're looking for a simpler workflow, check my <a target="_blank" rel="ugc" href="https://civarchive.com/articles/5366/comfyui-workflow">Main ComfyUI Workflow</a>. I also suggest to have a look at my <a target="_blank" rel="ugc" href="https://civarchive.com/articles/3160/prompt-notebook">Prompt Notebook</a> to better understand how I structure tags.If you have any comment or request, please feel free to share!<h2 id="features">Features</h2>Note: features in gray are in progress and need to migrate to v1.x<ul><li>Random procedural generation of prompts<ul><li>Main character (e.g. body, hair, eyes, tattos, piercing, horns, tail,…)</li><li>Attire (e.g. clothes, swimsuit, underwear, uniform, accessories.…)</li><li>Environment (e.g. place, daytime, nighttime, weather,…)</li><li>Action in scene (e.g. starting scene, sexual encounter, ending scene,…)</li><li>Predefined personas (demon, goblin, furry, slime, etc)</li></ul></li><li>One place to control all scenes parameters<ul><li>Seed, steps, CFG, image size, etc.</li><li>HighRes-Fix (2nd Pass)</li><li>LoRA stacker</li></ul></li><li>Keep control over scenes<ul><li>Re-generate one or many elements by changing its seed</li><li>Re-generate one or many scene by changing its seed</li><li>Overwrite and compose the final prompt with variables</li></ul></li><li>Scenes consistency<ul><li>Tags update dynamically according to the scene (e.g. <code>wet</code> is added if there's rain)</li><li>Attire state stay across scenes (if character lose clothes, it stay lost)</li><li>If clothes are torned, they stay so</li><li>Bondage ropes stay on character, with clothes</li></ul></li><li>Output<ul><li>Upscaling, pre-processing</li><li>Images scene merged into one</li><li>CivitAI metadatas &amp; workflow embedded</li></ul></li></ul><h2 id="setup">Setup</h2>Simply import the <code>workflow.json</code> file attached to this article in ComfyUI. You can also drag-and-drop the workflow image directly in the interface.I personally don't have a very powerful computer. For people in the same situation, check my environment on my <a target="_blank" rel="ugc" href="https://civarchive.com/articles/5366/comfyui-workflow">Main ComfyUI Workflow</a> article, I explain how I rent and setup machines on remote.<h3 id="models">Models</h3>In theory, the workflow can work with any models that use Danbooru-like tags. I personally use Illustrious/NoobAI/Pony-based models and mix with some anime-oriented LoRAs. Have a look at <a target="_blank" rel="ugc" href="https://civarchive.com/user/taches/images">my last images</a> metadatas if you're curious!<h3 id="custom-nodes">Custom nodes</h3>To achieve this workflow, I developed the comfyui-scene-composer extension. You can use as standalone in your own workflow. If you have trouble setting things up, check the <a target="_blank" rel="ugc" href="https://github.com/mus-taches/comfyui-scene-composer">repository</a>.

Scene Composer: Workflow

This version include the following changes:<ul><li>Denoise parameter is now globally changeable</li><li>LoRAs metadatas are displayed in output images</li><li>Seed for the prompt and the image generation are now clearly seperated, and can be changed globally or pro scene.</li><li>The scenes merged according to the ratio: vertical if it's landscape images, horizontal if it's portrait images.</li><li>Small breasts paizuri is improved (but still unstable)</li><li>Sometimes, clothes will be around character's waist (should be include in this-clothes)</li><li>If the character is gagged, no actions requiring an open mouth (e.g. blowjob) should appear</li></ul>

Files with SHA256: 5a0db3c830df2105ac89035cdf9442dc4a6c513635789ad97bb3bd9632541669