{"id":7255,"date":"2026-05-28T17:17:20","date_gmt":"2026-05-28T09:17:20","guid":{"rendered":"https:\/\/crepal.ai\/blog\/?p=7255"},"modified":"2026-05-28T17:17:23","modified_gmt":"2026-05-28T09:17:23","slug":"sulphur-2-comfyui","status":"publish","type":"post","link":"https:\/\/crepal.ai\/blog\/aivideo\/sulphur-2-comfyui\/","title":{"rendered":"How to Run Sulphur 2 in ComfyUI"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">Leo. I make content. I treat AI tools like LEGO. If I haven&#8217;t run a real project through it, I won&#8217;t write about it. I got Sulphur 2 running in ComfyUI last Tuesday. It took me about two and a half hours, and roughly 90 minutes of that was a VAE mismatch I should have caught in the first five minutes.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This is the sulphur 2 tutorial I wanted before I started. Everything in here is from actually running it \u2014 not from reading the model card and guessing. If something is uncertain or needs verification before you trust it, I&#8217;ll say so.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<h2 id=\"what-you-need\" class=\"wp-block-heading\">What You Need<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Before downloading anything, check your setup against this. Skipping this step is how people end up reinstalling halfway through.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Hardware floor:<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><tbody><tr><td class=\"has-text-align-center\" data-align=\"center\">Component<\/td><td class=\"has-text-align-center\" data-align=\"center\">Minimum<\/td><td class=\"has-text-align-center\" data-align=\"center\">Comfortable<\/td><\/tr><tr><td>VRAM<\/td><td>12 GB<\/td><td>16 GB+<\/td><\/tr><tr><td>System RAM<\/td><td>16 GB<\/td><td>32 GB<\/td><\/tr><tr><td>Storage (free)<\/td><td>25 GB<\/td><td>40 GB<\/td><\/tr><tr><td>GPU<\/td><td>RTX 3080 \/ equivalent<\/td><td>RTX 4080 \/ A100<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Twelve gigabytes is real but tight. At that floor, you&#8217;re running fp8 precision and keeping resolution at 512\u00d7512. At 16 GB you get longer clips and more headroom to push resolution without babysitting memory usage.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Software:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/github.com\/comfyanonymous\/ComfyUI\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">ComfyUI<\/a> \u2014 grab the latest stable commit. Video diffusion node support has changed a lot over the past year; anything more than six months old is likely to have compatibility gaps.<\/li>\n<\/ul>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-1 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"610\" height=\"237\" data-id=\"7260\" data-src=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-218.png\" alt=\"\" class=\"wp-image-7260 lazyload\" data-srcset=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-218.png 610w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-218-300x117.png 300w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-218-18x7.png 18w\" data-sizes=\"auto, (max-width: 610px) 100vw, 610px\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" style=\"--smush-placeholder-width: 610px; --smush-placeholder-aspect-ratio: 610\/237;\" \/><\/figure>\n<\/figure>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/github.com\/ltdrdata\/ComfyUI-Manager\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">ComfyUI-Manager<\/a> \u2014 you need this to install custom nodes without manually cloning repos one by one.<\/li>\n\n\n\n<li>Python 3.10 or 3.11. Python 3.12 has compatibility issues with several video nodes as of early 2026 \u2014 don&#8217;t use it here.<\/li>\n\n\n\n<li>CUDA 12.1 or later.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">If you&#8217;re already running LTX Video or Wan 2.1 locally, you probably have most of this in place. The sulphur 2 comfyui workflow shares a lot of infrastructure with LTX-based setups, so existing installs usually need minimal changes.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<h2 id=\"download-and-setup\" class=\"wp-block-heading\">Download and Setup<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Step 1: Get the checkpoint<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Go to the <a href=\"https:\/\/huggingface.co\/models?search=sulphur+2\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">Sulphur 2 model page on Hugging Face<\/a>. You&#8217;re looking for the main <code>.safetensors<\/code> file \u2014 around 13\u201316 GB depending on the variant posted. Community model naming isn&#8217;t always consistent, so read the model card before downloading to confirm you&#8217;re grabbing the inference checkpoint, not a LoRA or adapter file.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u26a0\ufe0f <em>This is a community release. Model card details and file names may change. Verify on the page before starting the download.<\/em><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Once downloaded, place it here:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>ComfyUI\/\n\u2514\u2500\u2500 models\/\n    \u2514\u2500\u2500 checkpoints\/\n        \u2514\u2500\u2500 sulphur2.safetensors<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Step 2: Install required custom nodes<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Open ComfyUI Manager (the &#8220;Manager&#8221; button in the top-right of your ComfyUI interface). Install these:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>ComfyUI-VideoHelperSuite<\/strong> \u2014 handles frame sequencing, video input, and video output. Without it, you have no clean way to export the generated clip.<\/li>\n\n\n\n<li><strong>ComfyUI-LTXVideo<\/strong> (or equivalent LTX-compatible sampler nodes) \u2014 Sulphur 2 is built on <a href=\"https:\/\/huggingface.co\/Lightricks\/LTX-Video\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">LTX Video 2.3<\/a>, which means it needs the matching inference architecture to run correctly. Standard KSampler alone won&#8217;t cut it.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">After installing both, do a full restart of ComfyUI \u2014 not a browser tab refresh, an actual restart of the server process. I know that sounds obvious, but I spent 15 minutes confused about missing nodes before I did this properly.<\/p>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-2 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" width=\"1024\" height=\"373\" data-id=\"7259\" data-src=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-217-1024x373.png\" alt=\"\" class=\"wp-image-7259 lazyload\" data-srcset=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-217-1024x373.png 1024w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-217-300x109.png 300w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-217-768x280.png 768w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-217-1536x559.png 1536w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-217-18x7.png 18w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-217.png 1571w\" data-sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" style=\"--smush-placeholder-width: 1024px; --smush-placeholder-aspect-ratio: 1024\/373;\" \/><\/figure>\n<\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Step 3: Download the VAE separately<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This is the step most sulphur 2 install guides skip, and it&#8217;s the one that gets people. The model uses the same VAE as LTX Video \u2014 it is not bundled inside the checkpoint file.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Download the VAE from the <a href=\"https:\/\/huggingface.co\/Lightricks\/LTX-Video\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">Lightricks LTX-Video repository on Hugging Face<\/a>. Look for the file labeled as the VAE or the combined model file from which VAE weights are extracted. Place it in:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>ComfyUI\/\n\u2514\u2500\u2500 models\/\n    \u2514\u2500\u2500 vae\/\n        \u2514\u2500\u2500 ltx_vae.safetensors<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\">Without the correct VAE loaded, you&#8217;ll get either black frames, visual noise, or an error during decode. The model loads fine \u2014 the failure only shows up at generation time, which makes it frustrating to diagnose.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<h2 id=\"comfyui-workflow\" class=\"wp-block-heading\">ComfyUI Workflow<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">There&#8217;s no single official sulphur 2 comfyui workflow file yet. What I&#8217;m giving you here is the node structure that worked consistently across my tests.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Core node chain:<\/strong><\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Load Checkpoint \u2192 sulphur2.safetensors\n        \u2193\nCLIP Text Encode (positive)\nCLIP Text Encode (negative)\n        \u2193\nLTX Video Sampler\n        \u2193\nVAE Decode\n        \u2193\nVideo Combine (VideoHelperSuite)\n        \u2193\nSave \/ Preview<\/code><\/pre>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Settings that worked in my runs:<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><tbody><tr><td class=\"has-text-align-center\" data-align=\"center\">Parameter<\/td><td class=\"has-text-align-center\" data-align=\"center\">Value<\/td><td class=\"has-text-align-center\" data-align=\"center\">Notes<\/td><\/tr><tr><td>Steps<\/td><td>25\u201335<\/td><td>Below 20 gets muddy fast<\/td><\/tr><tr><td>CFG scale<\/td><td>3.5\u20135.0<\/td><td>LTX-based models run lower than SDXL \u2014 don&#8217;t use 7+<\/td><\/tr><tr><td>Sampler<\/td><td>euler or dpmpp_2m<\/td><td>Both stable; euler is more predictable<\/td><\/tr><tr><td>Scheduler<\/td><td>sgm_uniform<\/td><td>Recommended for LTX-derived checkpoints<\/td><\/tr><tr><td>Resolution<\/td><td>512\u00d7768 or 768\u00d7512<\/td><td>Start here, push up once stable<\/td><\/tr><tr><td>Frames<\/td><td>25\u201349<\/td><td>Keep under 6 seconds until you know your VRAM ceiling<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">The CFG range is the biggest adjustment if you&#8217;re coming from image model workflows. I kept defaulting to 7.0 out of habit and getting overexposed, oversharpened outputs. Dropping to 4.0 fixed it. I ran that comparison six times to make sure it wasn&#8217;t prompt variance \u2014 it wasn&#8217;t.<\/p>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-3 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" width=\"1024\" height=\"474\" data-id=\"7258\" data-src=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-216-1024x474.png\" alt=\"\" class=\"wp-image-7258 lazyload\" data-srcset=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-216-1024x474.png 1024w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-216-300x139.png 300w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-216-768x356.png 768w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-216-1536x712.png 1536w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-216-18x8.png 18w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-216.png 1632w\" data-sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" style=\"--smush-placeholder-width: 1024px; --smush-placeholder-aspect-ratio: 1024\/474;\" \/><\/figure>\n<\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Prompt structure:<\/strong> Scene context before subject description consistently outperforms subject-first ordering on this model. &#8220;Sunlit office interior, woman in her 30s typing at a desk, natural movement&#8221; lands better than &#8220;woman typing at a desk, sunlit office.&#8221; I noticed this on my third test run and it held up across different subjects.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<h2 id=\"first-test\" class=\"wp-block-heading\">First Test<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Before running anything elaborate, use this prompt to confirm the setup is working:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Positive:<\/strong><code>Close-up of a ceramic coffee cup on a wooden table, morning light, subtle steam, cinematic<\/code><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Negative:<\/strong><code>blurry, distorted, watermark, text, cartoon<\/code><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Settings:<\/strong> 512\u00d7512 \u00b7 33 frames \u00b7 30 steps \u00b7 CFG 4.0<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This is intentionally low-complexity \u2014 no faces, minimal motion, controlled scene. If this doesn&#8217;t generate cleanly, the problem is in your node wiring or file paths, not your prompt. Figure it out here before burning time on harder shots.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">If it runs: generation takes about 4 minutes on an RTX 4080. Community reports for 12 GB cards with fp8 and <code>--lowvram<\/code> are landing at 8\u201312 minutes for similar settings. Both are in a usable range for testing.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Once the simple test passes, add a person to the scene and push resolution to 512\u00d7768. That&#8217;s where you&#8217;ll actually see what Sulphur 2 does differently from base LTX Video \u2014 the handling of skin tones and subtle facial movement is where it&#8217;s been getting attention.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<h2 id=\"common-errors\" class=\"wp-block-heading\">Common Errors<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">These are errors I hit directly, plus a few that showed up repeatedly in community threads.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&#8220;Model not found&#8221; or checkpoint missing from dropdown<\/strong> File is in the wrong folder or has an unexpected character in the filename. Check the exact path. ComfyUI is case-sensitive on Linux. Try clicking the &#8220;Refresh&#8221; button on the model loader node before assuming the file isn&#8217;t there.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&#8220;CUDA out of memory&#8221; at generation start<\/strong> You&#8217;re over your VRAM limit. In order: enable <code>--lowvram<\/code> in your launch arguments, switch to fp8 in the model loader, reduce frame count, reduce resolution. If you&#8217;re still hitting it at 512\u00d7512, fp8, with lowvram enabled, you&#8217;re below the hardware floor for this model.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>VAE decode produces black frames or visual noise<\/strong> Almost always the wrong VAE. Confirm you loaded the LTX-compatible VAE, not an SD 1.5 or SDXL VAE. They load without throwing an error \u2014 the failure only appears in the output, which is what makes it annoying to catch.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Custom nodes not showing up after install<\/strong> Full server restart, not a browser refresh. Most common cause, easiest fix.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>&#8220;NoneType has no attribute&#8221; error during sampling<\/strong> Node version mismatch. Go into ComfyUI Manager, update everything, restart. The error message is genuinely unhelpful \u2014 stale custom nodes cause this more often than anything else.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Motion flickering or jelly artifacts on moving subjects<\/strong> This is a model-level characteristic, not a setup error. Research on <a href=\"https:\/\/arxiv.org\/abs\/2501.06868\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">video diffusion model temporal consistency<\/a> shows frame-to-frame coherence across fast motion is still an open problem at this generation \u2014 Sulphur 2 doesn&#8217;t solve it. Mitigation: keep motion subtle, lower CFG slightly, add &#8220;smooth motion, consistent lighting&#8221; to your positive prompt.<\/p>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-4 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" width=\"1024\" height=\"450\" data-id=\"7257\" data-src=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-215-1024x450.png\" alt=\"\" class=\"wp-image-7257 lazyload\" data-srcset=\"https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-215-1024x450.png 1024w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-215-300x132.png 300w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-215-768x337.png 768w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-215-1536x674.png 1536w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-215-18x8.png 18w, https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/image-215.png 1836w\" data-sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" style=\"--smush-placeholder-width: 1024px; --smush-placeholder-aspect-ratio: 1024\/450;\" \/><\/figure>\n<\/figure>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<h2 id=\"faq\" class=\"wp-block-heading\">FAQ<\/h2>\n\n\n\n<h3 id=\"how-do-i-install-sulphur-2-in-comfyui\" class=\"wp-block-heading\">How do I install Sulphur 2 in ComfyUI?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Download the <code>.safetensors<\/code> checkpoint from Hugging Face and drop it in <code>ComfyUI\/models\/checkpoints\/<\/code>. Install ComfyUI-VideoHelperSuite and the LTX-compatible sampler nodes via ComfyUI Manager. Download the LTX Video VAE separately from the Lightricks repository and place it in <code>ComfyUI\/models\/vae\/<\/code>. Full server restarts after each node install \u2014 not just a page refresh. That covers the full sulphur 2 install sequence.<\/p>\n\n\n\n<h3 id=\"what-files-do-i-need-to-run-sulphur-2-locally\" class=\"wp-block-heading\">What files do I need to run Sulphur 2 locally?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Three things: the Sulphur 2 checkpoint (13\u201316 GB <code>.safetensors<\/code>), the LTX Video VAE file, and the ComfyUI custom nodes for video generation. The checkpoint does not bundle the VAE \u2014 you have to pull it separately from the Lightricks repository. A base ComfyUI install with standard image nodes is not enough for this workflow.<\/p>\n\n\n\n<h3 id=\"what-should-i-do-if-the-sulphur-2-comfyui-workflow-fails\" class=\"wp-block-heading\">What should I do if the Sulphur 2 ComfyUI workflow fails?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Start with the minimal test from the First Test section: 512\u00d7512, 33 frames, no faces, simple scene. If that fails, update all nodes in ComfyUI Manager and do a full restart. If the minimal test passes but complex prompts fail, the bottleneck is usually VRAM \u2014 enable <code>--lowvram<\/code>, reduce frames, drop resolution. For node-level errors not covered here, the <a href=\"https:\/\/github.com\/comfyanonymous\/ComfyUI\/issues\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">ComfyUI GitHub issues tracker<\/a> is the most reliable place to search by exact error message.<\/p>\n\n\n\n<h3 id=\"how-much-vram-does-sulphur-2-need\" class=\"wp-block-heading\">How much VRAM does Sulphur 2 need?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">12 GB is the practical minimum \u2014 fp8 precision, <code>--lowvram<\/code> flag, 512\u00d7512 resolution. At 16 GB you have room for 512\u00d7768 and longer clips without constant memory management. Below 12 GB, you&#8217;re not going to get usable output from this model regardless of settings.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<p class=\"wp-block-paragraph\">That covers the full how to use sulphur 2 setup from checkpoint to first output. The quality comparison \u2014 this model versus base LTX Video versus Wan 2.1 on matched prompts \u2014 is a separate post. Setup and evaluation are different questions, and combining them here would make this guide twice as long and half as useful.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">If you hit an error not in the Common Errors section, drop the exact message in the comments. I&#8217;ll either recognize it or test it and report back.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Tested on ComfyUI (early 2026 build), RTX 4080 16 GB, Windows 11 and Ubuntu 22.04. All linked resources were accessible at time of writing \u2014 community model pages can change without notice, so verify before downloading.<\/em><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-wide\" \/>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Previous Posts:<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-wp-embed is-provider-crepal-content-center wp-block-embed-crepal-content-center\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"wp-embedded-content\" data-secret=\"Mi8slT0l9F\"><a href=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-happyhorse-1-0-open-source\/\">Is HappyHorse 1.0 Open Source? What&#8217;s Actually Released<\/a><\/blockquote><iframe class=\"wp-embedded-content lazyload\" sandbox=\"allow-scripts\" security=\"restricted\" style=\"position: absolute; visibility: hidden;\" title=\"\u300a Is HappyHorse 1.0 Open Source? What\u2019s Actually Released \u300b\u2014CrePal Content Center\" data-src=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-happyhorse-1-0-open-source\/embed\/#?secret=PmRYcoOUQC#?secret=Mi8slT0l9F\" data-secret=\"Mi8slT0l9F\" width=\"600\" height=\"338\" frameborder=\"0\" marginwidth=\"0\" marginheight=\"0\" scrolling=\"no\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" data-load-mode=\"1\"><\/iframe>\n<\/div><\/figure>\n\n\n\n<figure class=\"wp-block-embed is-type-wp-embed is-provider-crepal-content-center wp-block-embed-crepal-content-center\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"wp-embedded-content\" data-secret=\"aftbthJWAg\"><a href=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-happyhorse-1-0-image-to-video\/\">HappyHorse 1.0 Image to Video: Full Guide &amp; Best Uses<\/a><\/blockquote><iframe class=\"wp-embedded-content lazyload\" sandbox=\"allow-scripts\" security=\"restricted\" style=\"position: absolute; visibility: hidden;\" title=\"\u300a HappyHorse 1.0 Image to Video: Full Guide &amp; Best Uses \u300b\u2014CrePal Content Center\" data-src=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-happyhorse-1-0-image-to-video\/embed\/#?secret=OcL31eMHm3#?secret=aftbthJWAg\" data-secret=\"aftbthJWAg\" width=\"600\" height=\"338\" frameborder=\"0\" marginwidth=\"0\" marginheight=\"0\" scrolling=\"no\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" data-load-mode=\"1\"><\/iframe>\n<\/div><\/figure>\n\n\n\n<figure class=\"wp-block-embed is-type-wp-embed is-provider-crepal-content-center wp-block-embed-crepal-content-center\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"wp-embedded-content\" data-secret=\"CiZLRGaHd0\"><a href=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-how-to-use-happyhorse-1-0\/\">How to Use HappyHorse 1.0: Step-by-Step Video Guide<\/a><\/blockquote><iframe class=\"wp-embedded-content lazyload\" sandbox=\"allow-scripts\" security=\"restricted\" style=\"position: absolute; visibility: hidden;\" title=\"\u300a How to Use HappyHorse 1.0: Step-by-Step Video Guide \u300b\u2014CrePal Content Center\" data-src=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-how-to-use-happyhorse-1-0\/embed\/#?secret=ulhNp2gYxY#?secret=CiZLRGaHd0\" data-secret=\"CiZLRGaHd0\" width=\"600\" height=\"338\" frameborder=\"0\" marginwidth=\"0\" marginheight=\"0\" scrolling=\"no\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" data-load-mode=\"1\"><\/iframe>\n<\/div><\/figure>\n\n\n\n<figure class=\"wp-block-embed is-type-wp-embed is-provider-crepal-content-center wp-block-embed-crepal-content-center\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"wp-embedded-content\" data-secret=\"5WKfY0L75w\"><a href=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-happyhorse-1-0-prompts\/\">HappyHorse 1.0 Prompts: Best Examples That Work<\/a><\/blockquote><iframe class=\"wp-embedded-content lazyload\" sandbox=\"allow-scripts\" security=\"restricted\" style=\"position: absolute; visibility: hidden;\" title=\"\u300a HappyHorse 1.0 Prompts: Best Examples That Work \u300b\u2014CrePal Content Center\" data-src=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-happyhorse-1-0-prompts\/embed\/#?secret=ivhqtuTXZU#?secret=5WKfY0L75w\" data-secret=\"5WKfY0L75w\" width=\"600\" height=\"338\" frameborder=\"0\" marginwidth=\"0\" marginheight=\"0\" scrolling=\"no\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" data-load-mode=\"1\"><\/iframe>\n<\/div><\/figure>\n\n\n\n<figure class=\"wp-block-embed is-type-wp-embed is-provider-crepal-content-center wp-block-embed-crepal-content-center\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"wp-embedded-content\" data-secret=\"LSyQ87pX8m\"><a href=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-qwen-image-to-video-tutorial\/\">Qwen Image to Video: How to Use It and What to Expect<\/a><\/blockquote><iframe class=\"wp-embedded-content lazyload\" sandbox=\"allow-scripts\" security=\"restricted\" style=\"position: absolute; visibility: hidden;\" title=\"\u300a Qwen Image to Video: How to Use It and What to Expect \u300b\u2014CrePal Content Center\" data-src=\"https:\/\/crepal.ai\/blog\/aivideo\/aivideo-qwen-image-to-video-tutorial\/embed\/#?secret=yVIc98CKAA#?secret=LSyQ87pX8m\" data-secret=\"LSyQ87pX8m\" width=\"600\" height=\"338\" frameborder=\"0\" marginwidth=\"0\" marginheight=\"0\" scrolling=\"no\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" data-load-mode=\"1\"><\/iframe>\n<\/div><\/figure>\n","protected":false},"excerpt":{"rendered":"<p>Leo. I make content. I treat AI tools like LEGO. If I haven&#8217;t run a real project through it, I won&#8217;t write about it. I got Sulphur 2 running in ComfyUI last Tuesday. It took me about two and a half hours, and roughly 90 minutes of that was a VAE mismatch I should have [&hellip;]<\/p>\n","protected":false},"author":10,"featured_media":7261,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_gspb_post_css":"","_uag_custom_page_level_css":"","footnotes":""},"categories":[8],"tags":[],"class_list":["post-7255","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-aivideo"],"blocksy_meta":[],"uagb_featured_image_src":{"full":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt.jpg",1376,768,false],"thumbnail":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt-150x150.jpg",150,150,true],"medium":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt-300x167.jpg",300,167,true],"medium_large":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt-768x429.jpg",768,429,true],"large":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt-1024x572.jpg",1024,572,true],"1536x1536":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt.jpg",1376,768,false],"2048x2048":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt.jpg",1376,768,false],"trp-custom-language-flag":["https:\/\/crepal.ai\/blog\/wp-content\/uploads\/2026\/05\/clean_Gemini_Generated_Image_nbytmdnbytmdnbyt-18x10.jpg",18,10,true]},"uagb_author_info":{"display_name":"Leo","author_link":"https:\/\/crepal.ai\/blog\/author\/leo\/"},"uagb_comment_info":0,"uagb_excerpt":"Leo. I make content. I treat AI tools like LEGO. If I haven&#8217;t run a real project through it, I won&#8217;t write about it. I got Sulphur 2 running in ComfyUI last Tuesday. It took me about two and a half hours, and roughly 90 minutes of that was a VAE mismatch I should have&hellip;","_links":{"self":[{"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/posts\/7255","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/users\/10"}],"replies":[{"embeddable":true,"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/comments?post=7255"}],"version-history":[{"count":1,"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/posts\/7255\/revisions"}],"predecessor-version":[{"id":7262,"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/posts\/7255\/revisions\/7262"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/media\/7261"}],"wp:attachment":[{"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/media?parent=7255"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/categories?post=7255"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/crepal.ai\/blog\/wp-json\/wp\/v2\/tags?post=7255"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}