HowTo (I2V but almost the same): 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
This workflow allow to create unlimited connected video by scenario description.
T2V video promt -> initial video
Precondition -> Constant text begin (atmosphere, actor description etc.)
String Selector -> Scenario (1 video = 1 line)
Postcondition -> Constant end of text (camera position, movements, mood etc.)
Description
The most powerfull workflow in this series.
Single T2V text
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)
FAQ
Comments (9)
do you give the loras link? thanks
I'm testing the new WF. Seems to be more optimized than that last one.
But, I don't for you, but for me, the WanImageToVideo node lags a lotttttt. Sometimes it takes a few minutes to unstuck on this node. And sometimes I need to close the CMD and restart Comfyui, because the flow is stuck on this node.
Of course, as with the previous WF, i immediately added, Resolution Master, sage attention, and tensorrt upscaling.
Sage attention not bout my Intel but...
Resolution master? Is that up/down scaler?
tensorrt upscaling - what is it?
And what did you use if not a WanImageToVideo? Any alternative?
@thesharque588 tensorrt is for Cuda RTX, it's an upscaler a lot faster than the normal upscaler.
Resolution Master, is a node to easily choose aspect ratio/resolution of the output image/video, with preset for almost all models. Try it, you will love it. I put it in all my WF now.
Ok, now the WanImageToVideo is stuck everytime i try to generate a video.
The first T2V scene work perfectely, but as soon as it start the I2V subsequent first scene, it's stuck on this node. I had the same issue with your previous WF.
I need to close the CMD and restart Comfyui to cancel the running generation.
And i only get this issue on your WF"s.
@Pat3dx I don't have cuda : ))) I have Intel i9 and that all : ))) it means NPU/XPU only
@Pat3dx WanImage take an image from For-Loop variable... looks like a bug in this component... Can you add node "Preview any" on this value?
@thesharque588 I check all my node conflict in manager, and even if i don't found a specific conflict for the WanImageToVideo node. I saw a lot of canflict with the Fantaisy talking nodes.
I deleted it, and it seems that's fixed the issue, hehehe!!!
@thesharque588 Add Resolution Master in your WF, you will love it :)
Finally managed to identify, download the main .gguf loras required for this to work, but as my 1st experience with .gguf, the .gguf quality is not good enough for real character lora so i replaced the six or so, .gguf nodes by the regular loader nodes for the .safetensors fp8 scaled variants and downloaded those instead. Real character lora identity is much better now. Also, I added Sage Attention and FP16 Accumulation nodes, so everything is 3,5x faster now. Replaced the standalone lora loaders in T2V section by power lora loader (rgthree). Upscaling large videos causes oom in the workflow so I will probably Topaz the best ones only. Thanks a lot for this genius extender workflow !