This is a clean and concise workflow for ComfyUI that allows you to generate longer videos by chaining together up to 4 separate clips.
V3.0 now includes SVI for better consistency in longer videos!
Note: Full instructions, advice, and links to models can be found in the note in the workflow itself.
This workflow is based on my clean 2+2[+2] lightning workflow that can be found here:
You can use this workflow to generate a single WAN 2.2 video or you can have it chain up to 3 additional videos based on the final frame of the previous clip. In this way you can create longer videos, with distinct changes in action or framing without needing to generate the whole thing in one go. Simply fine tune the first clip with the others disabled until you are happy. Then don't change that clip and move on to clip 2. ComfyUI won't spend time regenerating the earlier clips each time you want to tune the next one!
Please consider tagging this as a resource you used if you generate anything with it and upload it here. I'd love to see what you're making with the tool!
Description
Improved GGUF compatibility
Added SVI nodes
General tweaks
FAQ
Comments (35)
Hi, thanks for the workflow, works real nice. I was wondering, does integrating NAG in the workflow is a possibility ? I don't know if it's doable or not, maybe you've already tried ?
I hadn't come across or heard of NAG, I'll have to take a look and see! Thanks
I get more errors than I can count, for instance:
UnetLoaderGGUF 74:
- Value not in list: unet_name: 'DasiwaWAN22I2V14BTastysinV8_q8Low.gguf' not in []
It's a bit weird since I only use safetensors in this workflow, so it shouldn't complain about missing GGUF right? I notice it complains about a lot of different GGUF missing although I was careful to set it to NOT use GGUF.
This might be my bad workflow to be honest. You could try just deleting all the GGUF nodes and (the two model loaders and the CLIP loader), make sure the switch is set to False and try again?
Will version v3.0 work properly? Why is the node "Clip 1, Last Frame" not connected at all?
Prompt outputs failed validation: Required input is missing: images
SaveImage:
- Required input is missing: images
RIFE VFI:
- Required input is missing: frames
Hi there. I can't be sure of your exact issue, but it might be some missing custom nodes for the connectivity (the anything everywhere nodes) or it might be something has become disconnected/muted/bypassed.
All I can say is that the V3.0 should work in the state it's downloaded, as long as you have all the correct custom nodes installed.
I have same problem:(
i ony get a weird red ouput using clip 1 step 2, cllip 2 step 2, etc and [VideoHelperSuite] - WARNING - Output images were not of valid resolution and have had padding applied
So you should only use "step 2" if you use both "step 1" and "step 3" as well, but that might be what you meant?
@TheFatController well, everything work fine when i disable clip 1 step2, clip 2 step 2, clip 3step 2 clip 4 step 2, with any one of these enabled the final result is burn,everything getting red,something like that, i tried low and high, low only, high only but the final result its the same
@PadreF I'm afraid that without seeing your full workflow this is hard to diagnose! The step 2 is designed to work with the high noise models. I used to do a lot of generations with all three steps and from memory I would use a high CFG on step 1 (somethign like 3) then drop it to 1 for steps 2 and 3. Making sure you have the right loras loaded for step 2 is also important.
After a bit of fiddling I got it working, and I like this workflow. But, if I enable the "step 2" the resulting video is glitchy and oversampled looking. What am I missing? What is the point of the 3rd sampler? Also what does SVI do exactly?
The third sampler is actually the second sampler when the second one is disabled (how I run most of my generations). If you enable the second sampler, you need to make sure you have the right loras and CFG for it (the default as you download it are probably wrong!). So without seeing your exact settings I can't really advise much there. The best advice if you're unsure is to disable sampler 2 and just do a 2-step generation with the 1st and 3rd samplers.
The SVI helps with character consistency and loss of quality as you chain more clips together. With SVI the quality of multiple clips is improved and if your character moves out of shot, turns, face goes off - then when it/they come back they will still look like they did.
and where should I connect clip 1? it is not connected to you. The process isn't ready to go. I personally have no idea where to move it.
Everything should be connected properly with the "anything everywhere" nodes, so you won't see lines linking them. Do you have those custom nodes installed?
I have all the nodes installed, but when I launch it, I get an error on clip 1, which is disabled on your system. As a newbie, figuring out where to click it requires a lot of effort. I've seen this in other comments, too.
@bionovafood863 Clip 1 shouldn't be disabled I don't think. I usually disable clips 2, 3, and 4 at the start but clip 1 isn't.
There's a node called "fast bypasser" that can enable and disable various groups. Check that!
Yes, I have everything installed, but I couldn't get the workflow to start. I'm not that experienced with this stuff.
I'm using all your default models and the gguf flow, but every video is very slow motion. Is there a setting I'm missing or anything?
No amount of positive or negative prompting seems to fix it, sadly.
The first two things that come to mind are
1) Some Loras simply make the motion slow - I had one for "breast play" that would jarringly make a middle clip slow where the others were fine. So I'd check what loras you're using.
2) The "interpolation multiplier" can cause this. If it is too high compared to the frame rate of the output video then the whole video will show in slow motion. So you can try reducing this and see if that helps!
so much error. solved 5 error but still keep getting error. also force me to download unet even i using gguf, have to manual bypass. i give up this wf
Sorry you had so much trouble. For what it's worth the unet is gguf though, so that would make sense.
you lost me at the lora, dont know how to use that node
The Lora loader nodes? You simply add any loras you want to load by clicking "Add Lora" then choose the strength. There's one for each KSampler so you can add them independently wherever you like.
Which is the best lora for this ?
I have zero experience with comfyui and my computer is probably not up to snuff. Before using your workflow I have only generated some very low quality images because I've been stumbling around comfy over the last couple days like an overeager baboon. I found your workflow and followed your notes to the letter and was able to generate an I2V in 368s on my first attempt. I just wanted to say I appreciate all the effort you put into this! Thank you!
Glad it helped!
Hey hey I seem to have an issue where the second clip doesn t really have any motion like its not taking into consideration the prompt , the first one works great than i turn on the second clip enter a new prompt , generate but nothing I can see that there is some motion but barely any and defenitly not following the prompt for the clip 2 , I do have the step 2 and the upscale on off everywhere but I don t feel like that shuld be where the issue comes form. Any suggestion would be much appreciated , thanks !
Also the SVI lora has much better consistency and quality but not much motion , i tried it with a lightx 4step lora and motion was much better but not so much the quality , is there a way to have SVI with much more motion ?
Hey, I've run the exact workflow with all the nodes installed, no error, no nothing, while running the workflow. But why is the output only the 4 different PNG files? How to generate the long videos with it? Help with it?
Have you set the folders/filenames for the video save nodes?
@TheFatController yes, I do. the file name and folder set into the filename_prefix. the generated image got saved, but not the videos.
This is great but it crashes ComfyUI when it reaches the end .
Your workflow was great, but it's become completely unusable after the last update. I hope you'll update it soon...