{"id":55757,"date":"2026-08-15T09:41:03","date_gmt":"2026-08-15T01:41:03","guid":{"rendered":"https:\/\/www.1ai.net\/?p=55757"},"modified":"2026-08-11T09:28:24","modified_gmt":"2026-08-11T01:28:24","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%e5%87%ba%e5%9b%be%e8%b4%a8%e9%87%8f%e6%8f%90%e5%8d%87%ef%bc%88%e5%88%86%e8%be%a8%e7%8e%87-%e6%ad%a5%e6%95%b0","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/55757.html","title":{"rendered":"ComfyUI from installation entry to mastery, quality enhancement of the output map (resolution\/step\/sampler\/CFG\/high-resolution magnification)"},"content":{"rendered":"

\"0ceea579j00tjl02n00m7d000p000dyp\"<\/p>\n

Previous (Page 05)ComfyUI from installation to mastery, Workstream Practice Guide (read node chain + build + save reuse)<\/a> You've got a job lined up, and it's coming out. But does it feel like someone else's drawings are beautiful and clear, and your hair is gray, and your fingers are two more<\/p>\n

I JUST GOT OUT OF THE GRAPH, AND I JUST PULLED THE PARAMETERS -- STEP TO 100, CFG TO 15, AND THOUGHT, \"THE BIGGER THE BETTER, THE BETTER, THE MORE THE PLASTIC GOES. THEN I REALIZED THAT THE PAINT WASN'T STACKED WITH A STACK OF PARAMETERSWe'll move each to the right area<\/strong>I DON'T KNOW. THIS IS A GOOD ONE WITH A LIST OF THE 8G VISIBLE PRACTICAL PARAMETERS\u3002<\/p>\n

I. Resolution: multiple of 64 required<\/strong><\/p>\n

The different submersibles have its \"sizes recognized during training\":SD1.5 512 x 512, SDXL \/ Flux 1024 x 1024<\/strong>I DON'T KNOW. WHY MUST 64 BE MULTIPLE? BECAUSE THE VAE WILL CRUSH THE IMAGE INTO AN 8-SIZE SUBSPACE, NOT A MULTI-DIMENSIONAL DIMENSION, AND IT'LL BE CUT OFF. I'VE SEEN PEOPLE FILL 1000 X 1000, ODD EDGES, AND CHANGE TO 1024 IMMEDIATELY NORMAL\u3002<\/p>\n

8G VISIBLE ALARM:<\/strong>ON 4060, ABOUT 6.5GB BALANCE IS STABLE; UP TO 15362 IS EASY. THE 8G CARD STAYS IN 10242 AND WANTS TO USE THE \"HIGH-LEVEL MAGNIFICATION\" IN THE BACK, NOT HARD-LINE\u3002<\/p>\n

II. Number of steps: 20-30<\/strong><\/p>\n

THE NUMBER OF STEPS IS THE ITERATIVE NUMBER OF AI \"DRAW\" MAPS\u3002SDXL common 20-30 steps; Flux distillation only 8-12 steps<\/strong>I DON'T KNOW. I'VE MEASURED THAT SDXL FROM 20 TO 50 STEPS ALMOST THE SAME, AND TIME DOUBLES -- FROM 20 TO 30, IT DECREASES. NO, NO, NO, NO, NO\u3002<\/p>\n

iii. Sampler + Scheduler: Recognized gold combination<\/strong><\/p>\n\n\n\n\n\n\n
\n
Bottom Model<\/section>\n<\/td>\n
\n
Sample + Schedule<\/section>\n<\/td>\n
\n
clarification<\/section>\n<\/td>\n<\/tr>\n
SDXL<\/strong><\/td>\n\n
DPM++2M+Karras<\/section>\n<\/td>\n
\n
Gold standard, steady<\/section>\n<\/td>\n<\/tr>\n
SDXL, OUT<\/strong><\/td>\n\n
Euler a + Karras<\/section>\n<\/td>\n
\n
Try the hint word<\/section>\n<\/td>\n<\/tr>\n
Flux<\/strong><\/td>\n\n
Euler + simple<\/section>\n<\/td>\n
\n
Official formulation<\/section>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

IV. CFG: CONTROL OF THE EXTENT OF \u201cHEARING THE HINT\u201d<\/strong><\/p>\n

THE HIGHER THE CFG AI, THE MORE RIGOROUS IT IS; THE LOWER IT IS, THE FREER IT IS\u3002SDXL uses 5-8 (I usually use 7); Flux uses 1-4<\/strong>(Guidance around 3.5). Too high, exposure, plastic sense, hypocrisy -- I pulled 15, my face waxed; too low (<2) AI basically ignored you and painted. SDXL falls 5-8, most natural\u3002<\/p>\n

V. NEGATIVE: TELL AI NOT TO FIX A MOTH<\/strong><\/p>\n

\"Whatever you want,\" \"What you want,\" \"what you don't want.\" Half the quality is negative. I write in three categories:<\/p>\n\n\n\n\n\n\n
\n
form<\/section>\n<\/td>\n
\n
Word<\/section>\n<\/td>\n<\/tr>\n
\n
Quality class<\/section>\n<\/td>\n
\n
it's not like it's a bad idea<\/section>\n<\/td>\n<\/tr>\n
\n
Structure class<\/section>\n<\/td>\n
\n
extra fans, bad hands, deformed, mutagated limbs<\/section>\n<\/td>\n<\/tr>\n
\n
Interference class<\/section>\n<\/td>\n
\n
next, watermark, signature, username<\/section>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Seeds: fixed recurrence, redraw it<\/strong><\/p>\n

Seed is the \"random starting point.\"\u3002set rigidly in place<\/strong>(E. G. 12345) SAME PARAMETER 100% RECURRENCE; FILL\u00a0-1<\/strong>\u00a0(a) Every time a different image is explored\u00b11<\/strong>\u00a0Just a little light. I'm used to: randomly looking for images, fixed Seed, then fine-tweeting, 10 times more efficient than blind randomly\u3002<\/p>\n

VII. High-resolution Zooming: Small graphs come first, then magnify the details<\/strong><\/p>\n

If you want a big size, you're afraid of being exposed? Do not go directly to 15362, two steps with \"high-clean Zoom\": 1,10242 \u21922 Latent Upscale Zoom 1.5 to 2 times \u653e\u59273 \u21922 KSampler 2 fine (steps 10 to 20, denoise 0.3 to 0.6) 4 VAE Decode \u2192 Save. The details are better than the calculation of 50%+ directly from the Grand Chart (i.e. 05 \"Step A\")\u3002<\/p>\n

VIII. 8GB\/ 4060 LIST OF PRACTICAL PARAMETERS (COPY)<\/strong><\/p>\n\n\n\n\n\n\n\n\n\n\n
\n
parameter<\/section>\n<\/td>\n
\n
Value<\/section>\n<\/td>\n<\/tr>\n
\n
Resolution<\/section>\n<\/td>\n
\n
1024 x 1024 (not over)<\/section>\n<\/td>\n<\/tr>\n
\n
Steps<\/section>\n<\/td>\n
\n
25<\/section>\n<\/td>\n<\/tr>\n
\n
Sample + Schedule<\/section>\n<\/td>\n
\n
DPM++2M+Karras<\/section>\n<\/td>\n<\/tr>\n
\n
CFG<\/section>\n<\/td>\n
\n
7<\/section>\n<\/td>\n<\/tr>\n
\n
High Definition Zoom<\/section>\n<\/td>\n
\n
Latent Upscale 1.5x, phase II denoise 0.5<\/section>\n<\/td>\n<\/tr>\n
\n
Start Parameter<\/section>\n<\/td>\n
\n
\u2013medvram \u2013force-fp16<\/section>\n<\/td>\n<\/tr>\n
\n
other than<\/section>\n<\/td>\n
\n
watch = 1, no ControlNet, only 1 LoRA<\/section>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

This single SDXL diagram is about 8-12 seconds, and it's stable and stable. Flux Quantified Versions 8-12, CFG 1, Sample Euler\u3002<\/p>\n

It's all right, it's clear, but what happens when a bunch of red words pops upTitle 07<\/strong>Teach youYou know what I'm saying. Save yourself<\/strong>, the most common red word one by one\u3002<\/p>","protected":false},"excerpt":{"rendered":"

Last page, ComfyUI, from installation to mastery, Workstream Guide (read the node chain + build + save re-use), you put up the work stream, and the picture comes out. But does it feel like someone else's drawings are beautiful and clear, and your hair is gray, and your fingers are two more? I just got out of the graph, and I just pulled the parameters -- step to 100, CFG to 15, and thought, \"The bigger, the better, the better, the more plastic.\" It was later known that the paints were not stacked with stacks of parameters, and that each was moved to the appropriate compartment. This is a good one with a list of the 8G visible practical parameters. I. Resolution: must be 64 multiples with & 1 TP5T8221 in different submersibles<\/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-55757","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\/55757","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=55757"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/55757\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=55757"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=55757"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=55757"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=55757"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}