A workflow uses the latest SCAIL mode and 3D pose to transfer motions with perfect multi-character, spin movement, and complicated, swift move support.
All the model links are within the workflow.
Try the models online for free: https://www.runninghub.ai/post/2001100812002750466
Please check the video guide before using this workflow:
If you think the video is helpful, please like, share the video, and subscribe for more. I will update with more interesting things. Thank you.
Description
FAQ
Comments (22)
I keep getting error "rendernflposes index 0 is out of bounds for axis 0 with size 0
same
i have found a solution. This is an issue in the NLF node, when it cant detect a face or atleast 1 hand in even one frame, for example when back is turned or one hand goes outside frame. I have edited the NLF node to bypass detect check, therefore, it detects properly when the face and hand is visible and passes when not visible on those frames.
@Analyzer007 Hi, how do I do that exactly? I don't see "detect check" EDIT: Oops, nevermind I see now. Thank you
@mattoneplus12818 i have the edited py files, how do i send u?
Replace them inside NLF cutom_ nodes folder, one in its main folder and one in nlfposeextract folder
@Analyzer007 Hey there analyzer would you pls mind sending them to me i have this issue also ... igniteyourlimit_56986 (discord)
@Analyzer007 replacing these files did not help me. the same error with NLF
@mezonih75 interesting. make sure the nlf predict is osmething like 100 or 200 instead of -1
grok helped me edit the nodes .py and the render_nlf.py files and i've never gotten the error since. every video works fine (other than wan's inability to understand uncommon movements like yoga, acrobatics, or super fast actions)
Solution of error : "Index out of bounds"
Replace the bad code node files in Scail-Pose custom_node folder with these modified fixed python files.
This so freaking good, far better than steadydancer .
Thanks, Iv been trying to sort this for 2 days now
NLFPredict
The following operation failed in the TorchScript interpreter. Traceback of TorchScript (most recent call last): RuntimeError: invalid vector subscript.
Any one have any idea why i get this error ?
change value from -1 to basically anything like 50, 100 etc, if u get error again lower that value by half and try again, its a vram issue
@Analyzer007 thANKS
Hey guys. I unplugged the dw_pose from ref_dw_pose. It continues till it gets to WanVideo Model Loader, then it just crashes.
How to get the face consistency ?
Is there a way to put in an additional lora, like a jiggle lora for instance?
yes, connect them to the 'wanvideo lora select' node. you can use multiples. most of my examples have 3-5 wan2.1 loras. some wan2.2 loras work, but they can degrade the quality even at low weights
error (Download)Load NLF Model
Aside from changing the video duration amount, what else must be done to render over 65 seconds? every time I try 10 seconds I get an error no matter what.
lower res and fps. longest i can do is ~20 secs at 30fps 736x1280 on 32gb vram/64g ram. not sure if 65sec would be possible without serious quality loss. you likely need to split the clip, but be aware that the gen always drops the last few frames so split it so there is slight overlap