Rakan
Member
- Jul 19, 2017
- 297
- 267
- Thread starter BaasB
- Start date
Hazystomp
Member
- May 11, 2025
- 138
- 213
Slayerz
Active Member
- Aug 2, 2017
- 752
- 2,959
NTR.ai Love your animations. I also create animations in ComfyUI and I am wondering which model you are using for the animations.
Would it be Wan Image 2 Video with a Seamless loop workflow by using start image as last image of animation or simply ping-pong back and forth output?
And do you also use Vace with video pose reference? Your animations seems fluid, yet slowed down (interpolated?) I am trying to understand how you've made this massage animation:
Would it be Wan Image 2 Video with a Seamless loop workflow by using start image as last image of animation or simply ping-pong back and forth output?
And do you also use Vace with video pose reference? Your animations seems fluid, yet slowed down (interpolated?) I am trying to understand how you've made this massage animation:
Hazystomp
Member
- May 11, 2025
- 138
- 213
VoyeurRed
Member
- Mar 28, 2025
- 392
- 516
- Mar 23, 2025
- 150
- 248
Thanks! I use WAN I2V, and just simply ping-pong back and forth output. And no, I don't use VACE. But I do use interpolation.
8 more days for that. Boost your testosterone until then
Slayerz
Active Member
- Aug 2, 2017
- 752
- 2,959
Ahh I see. I used to also use the default ping-pong, but sometimes the results would look a bit off. You should check out this "Seamless loops" workflows on civitai, it's my go-to:
You must be registered to see the links
I am currently trying this new workflow using Lightx2V /SelfForcing for reduce render time and higher quality:
You must be registered to see the links
It is set to 480p, so I'm trying to test different loras/settings to have 720p going.
View attachment 235216_IN_00001.mp4
- Mar 23, 2025
- 150
- 248
Thanks for the info! I've actually never tried the seamless loop workflow because I figured adding more demands on the model would hurt quality. My assumption is same model, same knowledge base, but now it has to juggle more requirements, so it can't focus as much on the core output quality. That's based on what I picked up from Computer Vision stuff, though I could be totally off base.
I believe that there are decent loop video outputs out there, but I'd bet generally the success rate is lower compared to non-looped ones. That's what's kept me from experimenting with it. Have you noticed any performance differences between looped vs non-looped workflows?
For the second thing, I actually tested another workflow that had CausVid and Self-forcing in it, but it performed worse than my original setup. Thing is, there were other variables at play, not just those two components. Since they're just LoRAs anyway, I'll probably give them another shot in my original workflow and see how they perform in isolation.
darkforgestudios
Newbie
- Mar 12, 2025
- 26
- 229
Wow.
If you don't mind me asking... what I think it's more impressive on this game are the backgrounds. I wasn't ever able to pull out anything similar to this quality. How are you even doing this sorcery? If you don't mind me asking. Are you using a real image and rendering on top of it or are you really able to generate this type of quality with prompts?