{"id":55751,"date":"2026-08-11T09:15:16","date_gmt":"2026-08-11T01:15:16","guid":{"rendered":"https:\/\/www.1ai.net\/?p=55751"},"modified":"2026-08-10T17:36:44","modified_gmt":"2026-08-10T09:36:44","slug":"comfyui-%e4%bb%8e%e5%ae%89%e8%a3%85%e5%85%a5%e9%97%a8%e5%88%b0%e7%b2%be%e9%80%9a%ef%bc%8c%e7%95%8c%e9%9d%a2%e5%85%a8%e8%ae%a4%e8%af%86-%e4%bb%8e%e9%9b%b6%e6%89%8b%e6%90%93%e7%ac%ac%e4%b8%80%e5%bc%a0","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/55751.html","title":{"rendered":"ComfyUI from installation entry to mastery, interface to all + first image from scratch"},"content":{"rendered":"
\n
\n

\"b6463effj00tjjs0i00jd000os00dup\"<\/p>\n

Previous (Page 01)ComfyUI 5 Installation methods from installation entry to mastery, ComfyUI<\/a> Let's take.. ComfyUI<\/a> It's loaded, and it's the first one to verify. But to be honest, at that point you probably went \"by step, not knowing what you were doing.\" This is the one that's going to work -- you get a good look at the interface, you know what every square is, and you can put it out of a blank canvas\u3002<\/p>\n

Let's start with my embarrassment: I'm staring at the black box for half a day, afraid it's stuck, but it's loading the model. Comfyui starts slow and normal for the first time, don't panic\u3002<\/p>\n

Once you open it, you'll have three pieces<\/strong><\/p>\n

The first time I opened ComfyUI, I looked at the screens full of blocks and wires, the first reaction was, \"What kind of circuit.\" And then it turns out that it's better than a big frame like WebUIYou can see every step<\/strong>What a mistake to find\u3002<\/p>\n

1 Top menu bar\u3002<\/strong>The big button to the right, Queue Prompt, is \"start to produce\" and the shortcut Ctrl+Enter, remember, it will be used every day. On its left, the \"Load Default\" key loads official examples of work streams. If you put Manager on line 01, there'll be an extra Manager button in the menu bar. That's the entry for the plugs\u3002<\/p>\n

2 Intermediate large canvas\u3002<\/strong>The squares dragged and dragged are called \"nodes\" and the lines connected between them are called \"data flows\". Each node does a specific thing: read a model, write a hint, take a sample ... Think of it as a plant waterline, raw materials go in from the left, finished products come out from the right, and the line is a material transfer belt\u3002<\/p>\n

3 Right Parameter Panel\u3002<\/strong>Click a node to show its parameters on the right: the model drops down, the hint input box, the step slider, all of which is changed here\u3002<\/p>\n

An operation to remember:<\/strong>It's blank in the canvasdouble click<\/strong>, a search box will pop up, and a node will be added by entering a node. 80% of starters rely on it. I just used the \"plus nodes\" button to fill the world, and then I realized that double-clicking was the way to go \u2014 a technique that nobody would tell me about, and it would get stuck for half a day. The interface is all English. Title 01, part X, teaches how to cut Chinese. Just put it in the menu\u3002<\/p>\n

ii. Make your first drawing<\/strong><\/p>\n

We'll use the official default stream, run and identify\u3002<\/p>\n

Step 1: Load the default workflow\u3002<\/strong>Pointtop \"Load Default\" (some versions in the WorkFlow menu point Open selection). Pow, there's an already connected node on the canvas, which is the official \"Standard Process of the Ventural Chart\"\u3002<\/p>\n

Step 2: Recognize this node chain\u3002<\/strong>This is the one from the left to the right:<\/p>\n

Load Checkpoint
\n\u2022 CLIP Text Encode is positive\/negative (writing hints)
\nEmtty Late Image
\n\u2192 KSampler
\nVAE Decode
\nSave Image<\/p>\n

This is the key to understanding ComfyUI:<\/strong><\/p>\n\n\n\n\n\n\n\n\n\n
\n
node<\/section>\n<\/td>\n
\n
What's it doing<\/section>\n<\/td>\n<\/tr>\n
Load Checkpoint<\/strong><\/td>\n\n
Load the base model, the base of the map<\/section>\n<\/td>\n<\/tr>\n
CLIP Text Encode<\/strong><\/td>\n\n
TRANSLATION OF TEXT HINTS INTO A VECTOR UNDERSTOOD BY AI (UP-AND-DOWN NEGATIVE)<\/section>\n<\/td>\n<\/tr>\n
Emtty Late Image<\/strong><\/td>\n\n
Set a blank canvas size in the subspace<\/section>\n<\/td>\n<\/tr>\n
KSampler<\/strong><\/td>\n\n
Core engine, step by step by parameter in subspace<\/section>\n<\/td>\n<\/tr>\n
VAE Decode<\/strong><\/td>\n\n
Decoding subspace data to images that adults can see<\/section>\n<\/td>\n<\/tr>\n
Save Image<\/strong><\/td>\n\n
save disk, drop output folder<\/section>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Step three: Bottom model\u3002<\/strong>Point \"Load Checkpoint\", right under ckpt_name select a model you download. Newcomers recommend SDXL (e.g. RealVisXL), 4060 running steady. Is it empty? That means you haven't done the model yet\u3002<\/p>\n

Step four: Write a hint\u3002<\/strong>Point \"CLIP Text Encode\" node, text box entered: a beautiful girl, sunny day, detailed face. Point negative to that, enter: bad hands, extra finers, blurry, low quality. The negative hint is, \"Tell AI not to fix these moths.\"\u3002<\/p>\n

Step 5: Size\u3002<\/strong>Point \"Empty Late Image\", width\/head set 1024 x 1024\u3002SDXL and Flux use 1024 x 1024, only old SD1.5 uses 512 x 512\u3002<\/strong>4060 If it's tight, the first 1024 will drop to 768 x 768\u3002<\/p>\n

Step 6: The sampler stays put\u3002<\/strong>Dot \"KSampler\", steps (steps) 20, cfg 7, sampler select euler or dpmpp_2m, scheduler select normal, seed set -1 (random). Use the default, run and play\u3002<\/p>\n

Step 7: Run<\/strong>Dot top \"Queue Prompt\" (or press Ctrl+Enter). After the bottom progress bar, the Save Image node will pop up and the picture will be stored in the output folder. I waited for almost a minute for my first drawing, and I thought it was stuck again, and the results came out -- normal, big models slow\u3002<\/p>\n

Well, you know how to get out this time, not blind spot\u3002<\/strong><\/p>\n

I'll save it for you<\/strong><\/p>\n

For the first time, the proposal is immediately saved. Points Menu WorkFlow Save (or Export), generates a .json file. Double-click this json, work-stream one key back. Now it's my habit to have a json on every fine-tuned map, a copy of which is double the efficiency. It's a lot more than every single time\u3002<\/p>\n

I'll save you two hours<\/strong><\/p>\n\n\n\n\n\n\n\n
\n
phenomena<\/section>\n<\/td>\n
\n
How<\/section>\n<\/td>\n<\/tr>\n
\n
Load Checkpoint<\/section>\n<\/td>\n
\n
No model. Next SDXL, put models\/checkpoints, refresh<\/section>\n<\/td>\n<\/tr>\n
\n
Could not double-click the search box<\/section>\n<\/td>\n
\n
Make sure you double-click the canvas blank, not the node<\/section>\n<\/td>\n<\/tr>\n
\n
Wrong red word for drawings<\/section>\n<\/td>\n
\n
Read last line of red; read page 06 in detail<\/section>\n<\/td>\n<\/tr>\n
\n
4060, it's bursting<\/section>\n<\/td>\n
\n
USE SDXL+1024 FIRST, OR A QUANTITATIVE VERSION OF THE GGUF (17TH EDITION)<\/section>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Now you can run ComfyUI up and read the node chain. But there is only one bottom model, there is no plug-in, and there are only a few things that can be done. I'll teach you the next oneWhere does the model file go, where it goes, how does it fit with Manager<\/strong>It's a key step in getting ComfyUI from running to working\u3002<\/p>\n<\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"

Last page (Page 01) ComfyUI was installed from the first door of installation to the first of the five ways that ComfyUI was installed. But to be honest, at that point you probably went \"by step, not knowing what you were doing.\" This is the one that's going to work -- you get a good look at the interface, you know what every square is, and you can put it out of a blank canvas. Let's start with my embarrassment: I'm staring at the black box for half a day, afraid it's stuck, but it's loading the model. Comfyui starts slow and normal for the first time, don't panic. One, don't snuggle, just three things<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[149,144],"tags":[1989,6276,4749],"collection":[],"class_list":["post-55751","post","type-post","status-publish","format-standard","hentry","category-jiaocheng","category-baike","tag-comfyui"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/55751","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=55751"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/55751\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=55751"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=55751"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=55751"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=55751"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}