HowTo: https://civarchive.com/articles/19108
UPD: New version of Loop + Scenario + Loras
The most powerfull workflow in this series, highly recomend update to this one.
Single precondition text
Single postcondition text
Unlimited scenario steps (5 by default but you can add)
Unlimited lora for all (1 by default)
Unlimited SEPARATE Lora per step (1 by default)
Color correction
UPD: New version of Loop + Scenario + Color is here!
Separate precondition
One-line scenario
Separate postcondition
Color correction is works now : )
Updated version of https://civarchive.com/models/1826257
Now you can use Scenario with different prompt for each step.
This workflow contain For-Loop so you can create as much step as you wish just copy-past Prompt for each step and change what you need.
Do not forget to change precondition (for ex. If you undress wooman and take off sweater do not desribe next step as "weared in sweater" - it confuse model)
Lighting: https://huggingface.co/Kijai/WanVideo_comfy/tree/main/Lightx2v
FastVideo: https://huggingface.co/Kijai/WanVideo_comfy/tree/main/FastWan
Wan2.2: https://huggingface.co/QuantStack/Wan2.2-I2V-A14B-GGUF/tree/main
Description
The most powerfull workflow in this series.
Single precondition text
Single postcondition text
Unlimited scenario steps (5 by default but you can add)
Unlimited lora for all (1 by default)
Unlimited SEPARATE Lora per step (1 by default)
Color correction
FAQ
Comments (37)
What an amazing piece of work. This has been a big problem for wan2.2 users and I am very impressed you have fixed it with this fantastic workflow. Massive respect to you.
Next rime I'll update my T2V version... But TI2V-5B is still kill me, nothing better thatn trash : )
nice job! as a noob, i don't understand how to make use of pre-condition, string selector and post-condition. anybody kindly give me a detailed explanation?
Precondition - what you see on the picture (back ground, persons, something static)
Postcondition - static addon (expression, camera stand/move, slow/fast movements etc)
Precondition + step 0 line + postcondition
Precondition + step 1 line + postcondition
I'ts just to not write "Busy street background, low light..." each time : )
Download workflow - you can see an example there
@thesharque588 how do i add more steps, and a separate prompt for each?
@favas644 https://civitai.com/articles/19108 check this out. 1 line = 1 step scenarion
@favas644 hm... Why I see 3 similar comment... Anyway : ) Check this HOWTO https://civitai.com/articles/19108
@thesharque588 Great, i was missing out the part of the "total step count" and the thing about separate loras for each step. That's why my generation was unconsistent, now everything works correctly! Just another question: is there a way to keep the completed steps, for example when i already have some nice steps, but something goes wrong later? Or do i have to redo all the job from the beginning?
@favas644 Right now only redo all... but I also think about it, and it looks required
Could you explain how to add more scenario steps to the workflow please, and if anything else needs to be adjusted? I ran into an error for NodeInputError...(says it needs input value5, but there is no input to that node at all). I tried increasing the steps in the noise models to 10 and adjusted the start_at_step values
Nooo no no.. There is another steps - it's a "render steps" Let me write this in description
Check this out, hope my english do not destroy sence : ))) https://civitai.com/articles/19108
Thank you for your work on this! When using your workflow, i get this error: Node 78.0.0.3.0.0.132 says it needs input value1, but there is no input to that node at al
Any ideas would be great, thanks!
Looks like you have no init image? Or you bypass color correction?
Very clever idea, and your layout is superb. No superfluous set/gets, every section is clearly labeled. Simple concat, indexed for the for loop... chef's kiss. Beautiful. Most other examples of this concept I've seen are bloated disasters that don't even seem like they were intended to be used, but rather to be clicked on. I'm not quite ready to surrender my editorial choices to automation ( though not for lack of trying), but there are a bunch of great techniques here to steal. Takes me twice as many nodes to do some things you do here...embarrassing. And I highly encourage every effort to end the plague of 3 second videos. The internet is almost full, I don't know how much more it can take. This site is already panting and groaning like a dying yak.
Thank you a lot, steal everything you wish 😂 in my case I just can't create big and long video, but I like to experiment.
I don't understand. scenario+loras workflow doesn't look anything like what you describe in HOWTO(but previous one - scenario+clip+color actually does). Did you accidentally put wrong workflow in the archive?
Nope I've check Dwonload and it is Wan2.2 + Lora (of course with color correction and other previous features)
Can I replace ImpactStringSelector by other prompt descriptor ? I have issue with Comfyui-impact Nodes
This node take 1 line by index so any "String selector" is good enough.
Highly recommend adding some Clear VRAM Used nodes into this workflow. I added them as outputs for Load Image, Negative Prompt, Model Sampling low and high, Low noise Ksampler, VAE Decode, and One step preview. This let me add an Interpolate node for each step and improved performance so I was reliably able to do 5 total steps at .75 megapixels with 97 frames per step (4090 with 24gb vram).
Hi! Thank you for the advice! What "Clear VRAM" did you used? Also about "interpolate node" do you interpolate frame or latency? Can you upload your WF somewhere and I steal this : )))
@thesharque588 I used the Clean VRAM Used node from the "EasyUse" custom node pack. For interpolation I use the RIFE VFI node from ComfyUI-Frame-Interpolation custom node pack and I usually just do 2x frames to smooth things out (usually important to double the FPS of the final video unless you want slow motion).
My workflows are a complete hodgepodge mess that really only make sense to me. I'm still tweaking things a lot as I figure out what works best for me, but once I settle on some relatively stable workflows, I'll try to upload them.
I will add this, thank you! But (in case of experiment) can you check this flags?
--cache-classic OR --cache-none - it looks like you don't need a cache...
Somebody has to find a way to keep the good segments and redo the bad ones only. Redoing the whole job when something goes wrong is not just time consuming, it also generates garbage data that will affect frther generations. I've been doing research, but i don't have enough tech skills to do that, unfortunately.
Why is there flicker in the first few frames from the input image to the video output? I can't seem to fix this. I get this with all for loop workflows, really annoying. Anyone know how to fix?
Remove... Just use Split node with param 2 and use part B - usually it's a bug of Scheduller / Sampler. You can try other (I use dpmpp_2m_sde + sgm_uniform for WAN and SDXL)
@thesharque588 that actually fixed it. issue was with res_2m + bongtangent
Thank you so much. This is the best workflow for creating long animations with a story. Simple and effective!
Works great on a 3060 with 12GB + 32GB of RAM.
Now I'm dreaming of a new, more powerful video card=)
Does anyone know why the first step is perfect but the second scene speeds up?
Update comfy and plugins (extension) It is because of VHS. Saw similar. Also It looks like 37, 81, 120 frames generate different FPS : )) in my case 37 is faster (about 16 FPS) than 120 (about 24 FPS). Also come Lora slow down video to 24FPS
the best workflow that works LOL!
And you put links to the models - Awesome!!!
This should be a standard on Civitai - embedding in the videos that can be duplicated
Hi,
Great workflow, thanks!
If I want a video loop without changing the style of my image, is that possible (I want to create a wallpaper while keeping the position of my character intact/fixed)?
Actually IDK - try to use empty prompts
@thesharque588 Alright, thanks I will try.