\n| Video input (Turson video reference)<\/td>\n | Same memory<\/td>\n | The decoded reference video will also buffer the entire frame<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n And it's not just the \"out of video\" step. You use itVideo input<\/strong>The decoded reference video also buffers all frames into the memory, with the same size\u3002<\/p>\nii. v0.29 changed: video flow code<\/strong><\/p>\nComfyUI\u00a0v. 0.29.0<\/strong>2026-07-29 published) In the chapter of Performance & Stability, this section has been specially adapted. Official PR describes it as:<\/p>\nVideo transport (PR #14813)<\/p>\n It's just a translation:The video transcoder was converted to a fluid treatment - each frame is no longer buffered into memory\u3002<\/strong>It's called \"Salvaging All Frames and Writing Discs\", it's now \"Generating One Frame, Writing One Frame, Release One Frame\" and the memory always leaves only a small amount of frame currently being processed. Long video, high resolution benefits the most, as memory occupancy no longer surges with frames\u3002<\/p>\nalong with v0.29, there are two other things related to memory\/ resource that deserve to be mentioned:<\/p>\n 1.\u00a0BACKSTAGE MEMORY LIMIT DEFAULT REFERS TO 128GB<\/strong>(PR #15068) The upper limit of the CofyUI backstage occupancy system is raised, and large memory machines are less likely to be crushed by it\u3002
\n2.\u00a0Comfy Kitchen More Int8 Optimise<\/strong>(PR #14980): Int8 quantifying this road earlier (v 0.27 enters the main line), v0.29 is continuing to grindIt's not new<\/strong>\u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002 \u3002<\/p>\nIntegrity:<\/strong>IT'S CALLED \"VIDEO FLOW-CLASSED MEMORY\" WHICH I SAW AS A MECHANISM IMPROVEMENT FROM THE OFFICIAL PR DESCRIPTIONWe're not running on the card<\/strong>Super-long video was compared back and forth, with no public figures for the specific drop. But in principle, long video, high-resolution scenes benefit most; you can feel it when you upgrade\u3002<\/p>\niii. Easy to mix: This is saving memory, not saving memory<\/strong><\/p>\nPart 18 What about ComfyUI, from installation to mastery? A list of downgrades by priority<\/a> IT'S VRAM, IT'S MEMORY. THEY'RE NOT THE SAME:<\/p>\n \n\n\n| Symptom<\/td>\n | Gene<\/td>\n | Which one<\/td>\n<\/tr>\n | \n| CUDA out of memory<\/td>\n | MODEL \/ VAE CANNOT LOAD VISIBLE<\/td>\n | List of downgrades<\/td>\n<\/tr>\n | \n| Process systematically killed \/ card killed<\/td>\n | Frame buffer full memory<\/td>\n | title (iv. 0.29.2)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Video generation is a double-pressure \"Standing + Memory\" scene. A lot of people just show up, and they fall, and it's the half of the memory\u3002<\/p>\n clean loop: upgrade to v.29.2 and run a long video (do it now)<\/p>\n Let's get to the point. This one's real. After you've got another video, \"Doesn't fail under a long video,\" you'll be able to confirm that the current code actually works\u3002<\/p>\n Preconditions<\/strong>: You've been running videos to generate workflows (Van, HunyuanVideo, or trinking nodes). Dropped: Basic Textile Chain (Page 05) + Video-related Node (Load Video \/ VAE Encode \/ Decode + Video Encoding Node)\u3002
\nTarget<\/strong>: Raise ComfyUI to v.29.2 and rerun a video of \"intentionally lengthening\" with a more stable memory occupancy and no further collapse\u3002<\/p>\nStep (sight, don't jump):
\n1.\u00a0Update ComfyUI<\/strong>: Desktop Dot \"Check Update\" upgrade; Git release restarts after ComfyUI directory guit pull\u3002
\n2.\u00a0Confirm Version<\/strong>: Menu Help \u2192 About, see version number V0.29.2 (2026-07-31 latest stable version)\u3002
\n3.\u00a0Rerun Video Workstream<\/strong>, THIS TIME INTENTIONALLY PULLS OUT MEMORY PRESSURE FOR A LONG TIME OR AT A RESOLUTION (5 SECONDS \u219210 SECONDS, OR 480 P \u2192720P)\u3002
\n4.\u00a0Look at the memory curve<\/strong>: With the task manager on, the old version will increase steeply at the end of the generation, and the new version should be much more flat and no longer flush\u3002<\/p>\nCheckpoint<\/strong>
\n1 About displays v.29.2; long video runs and dies
\n2 Common error = an updated plugin with an error \"Node Not Found\" plugin needs to be reloaded \/ updated (updated in Manager); or an additional 18 downgrade list is not visible enough
\n3 still collapses, either memory or display double pressure, to see part 18 in the half; memory still closes to start-up parameters \u2013cache-none
\n4 Video came out but the resolution\/ step question is not a transcode\u3002<\/p>\nJob (chosen): The same video, one before and after the upgrade, tells me what the difference is in the memory peak -- I'm curious to know how much you can save\u3002<\/p>\n iv. Parameter speed sheets (video memory relevant, copying)<\/strong><\/p>\n\n\n\n| take<\/td>\n | Recommendations<\/td>\n | Remark<\/td>\n<\/tr>\n | \n| SHORT CLIP (<5 SECONDS 480P)<\/td>\n | It doesn't make any difference<\/td>\n | Less frame, less pressure<\/td>\n<\/tr>\n | \n| Long Snippet (>10 seconds)<\/td>\n | i have to go to v.29.2<\/td>\n | Fluid transcode yields the most<\/td>\n<\/tr>\n | \n| HIGH RESOLUTION (720 P+)<\/td>\n | ig v0.29.2 + 18 visible decline<\/td>\n | Memory + Visible Double Pressure<\/td>\n<\/tr>\n | \n| Video input (Turson video reference)<\/td>\n | l v.29.2<\/td>\n | It's also good to decode<\/td>\n<\/tr>\n | \n| Memory remains tight (extreme environment)<\/td>\n | add -cache-none<\/td>\n | Uncut intermediate results, save memory exchange speed<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n v. Contrast of start-up parameters (RAM, copying)<\/strong><\/p>\n\n\n\n| parameter<\/td>\n | corresponds English -ity, -ism, -ization<\/td>\n | For what<\/td>\n<\/tr>\n | \n| (default)<\/td>\n | v0.29 streaming code is open<\/td>\n | up to v. 0.29.2<\/td>\n<\/tr>\n | \n| \u2013cache-none<\/td>\n | No Cache Node Results<\/td>\n | The memory is on the bottom<\/td>\n<\/tr>\n | \n| \u2013reserve-vram 1.5<\/td>\n | Save System Visibility<\/td>\n | When double pressure is visible \/ memory (see under 18)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n One sentence:only upgrades to v. 0.29.2, transcode automatic flow; until memory limits add \u2013cache-none\u3002<\/strong><\/p>\nVI. 4060 \/ 8G SPECIALIZED & I STEPPED ON THE PIT<\/strong><\/p>\n4060 THIS CARD, VISIBLE 8G, MEMORY USUALLY 16~32G. VIDEO GENERATION IS THE DOUBLE PRESSURE OF THE \"DIGNO REPOSITORY MODEL + MEMORY CACHE FRAME\": TO SEE PART 18 FOR THE HALF, THE HALF FOR THE MEMORY\u3002<\/p>\n Trail experience:<\/strong>I ran a 10-second graphic video, only adjusted the visual parameters, the memory was undone, and the frame buffer was 32G full, the system killed the process. Then I went up v. 0.29, and the resolution dropped from 1080 P to 720 P, and the same thing went through. So..Video blast memory, first version, then resolution<\/strong>Don't die on display\u3002<\/p>","protected":false},"excerpt":{"rendered":"In the previous 18 articles, I told you about the machines, the Ventura, the TowerNet, LoRA, and a list of downgrades for you. And as soon as the article came out, the first person in the comment section said, \"What's the use of video-generated RAM? I was a little bit surprised because he was right. For the first time in my life, I used ComfyUI to make a video, a five-second little clip, to run half the whole software card. Cut to the task manager -- 32G memory is full, the system kills the process. I thought it was a graphic card that didn't work, and it was a half-day display of parameters in line 18 that came back: RAM, not VRA<\/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-55814","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\/55814","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=55814"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/55814\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=55814"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=55814"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=55814"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=55814"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}} | | | |
![\"68bc142cj00tjlfsu00m3d000os00e4p\"](\"https:\/\/www.1ai.net\/wp-content\/uploads\/2026\/08\/68bc142cj00tjlfsu00m3d000os00e4p.jpg\" "\\"68bc142cj00tjlfsu00m3d000os00e4p\\"")
<\/p>\n