Two ComfyUI Additions Useful for Stable Diffusion AI Art
<p>With SDXL, there’s a lot going on in the ComfyUI world. There are many nodes and complex workflows.</p>
<p>There are two things I’ve added to my basic SDXL workflow lately that have improved my results:</p>
<ol>
<li>The “Wrong” SDXL LoRA. Less-bad hands.</li>
<li>The Ultimate SD Upscaler custom node.</li>
</ol>
<p>I have a link to a PNG image with my embedded flow later in this article.</p>
<h1>A Wrong to Make It Right</h1>
<p>The Wrong LoRA can be found here: <a href="https://civitai.com/models/128708/sdxl-wrong-lora" rel="noopener ugc nofollow" target="_blank">SDXL Wrong LoRA — v1.0-Diffusers | Stable Diffusion LoRA | Civitai</a></p>
<p>There’s a good write up about the process of creating this LoRA here: <a href="https://minimaxir.com/2023/08/stable-diffusion-xl-wrong/" rel="noopener ugc nofollow" target="_blank">I Made Stable Diffusion XL Smarter by Finetuning it on Bad AI-Generated Images | Max Woolf’s Blog (minimaxir.com)</a></p>
<p>You add the LoRA into your model flow and then add the word “wrong” to your negative prompt.</p>
<p>Snippet of my flow:</p>
<p><img alt="" src="https://miro.medium.com/v2/resize:fit:700/1*Om9oXTTK2mthi4GiFba9Jg.png" style="height:502px; width:700px" /></p>
<p>The sdxl-wrong-lora LoRA Loaded in ComfyUI</p>
<p>And just to be clear: I’m adding three LoRAs to my model flow here: the “Wrong” LoRA we’re discussing, one to make things look more artful, and the Stability AI offset LoRA. You can just do the “Wrong” loRA. The last in the flow chain gets connected to everything else.</p>
<p><a href="https://ericri.medium.com/two-comfyui-additions-useful-for-stable-diffusion-ai-art-49d4e21f4a2f"><strong>Read More</strong></a></p>