Simple workflow for loop i2v animation.
Missing custom nodes can be installed by ComfyUI-Manager.
Based on parts from other workflows:
Wan 2.1 seamless loop workflow (i2v) - v1.0 | Wan Video Workflows | Civitai
Description
Using both 2.2 models and two identical images generates almost still image.
So here is mixed workflow. WAN 2.1 used as high model and WAN 2.2 as low model.
FAQ
Comments (18)
WanFirstLastFrameToVideo
module 'tensorflow' has no attribute 'Tensor'
Can please someone eexplain why all of generation get strange rainbow effect on image? I try resolution 672x1024, 3 sec video. Only using loras for Wan 2.2 LightSpeed lora (H&L) and NSFW general (H&L).
What model are you using for first sampler? You need 2.1 loras for 2.1 model. Last workflow generates movement with wan 2.1 model and wan 2.2 for second sampler.
@Discocat oh, i understand now, i was using for both samplers wan 2.2 model. So i need to put wan 2.1 model in first sampler and use for first only 2.1 lora, am i right?
To not waste your time later: why we can't use wan 2.2 overall? I'm new with video geenration and trying to understand)
@legolas8456 Wan 2.2 generates almost still image (without any motion) if you are using same start and end image. In this workflow wan 2.1 generates movement and the wan 2.2 generates details.
@Discocat thx, i have one more question)
I used another workflow for loops before yours, and when i changed amount of frames i press RUN again - its not start generation from beginning, its start after all frames generated and just skip them in final result.
How can i do the same in yours workflow? Sometimes i need to try skip some frames for several times and wait every time untill its generate again from beginning is very annoying...
Sorry for my english, its can be really bad, but i hope you can understand me)
@legolas8456 There is node "Seed (Randomizer)". With fixed seed value and same prompt ComfyUI should skip generation and use same result.
@Discocat i did "New fixed random" but its anyway start from beginning...
hey can i ask, did you find any way to control speaking/talking animation? I tried evry diffrent workflow and i tried to modify many myself. Diffrent models, loras. Nothing work.
I can write 50 negatives for talking and 50 postivies for closed mouth. Characters still want to sing even if its I2v and starting image was with closed mouth.
Very nice workflow btw,
i am using negative prompt "speaking, talking, open mouth", but it doesn't always work
I guess it’s because talking animations give generated characters that extra sprinkle of humanity, like a digital puppet suddenly deciding it’s alive. :))
Guys, how are you fixing eyes and other artifacts? I was trying to add more steps and change loras, but they all time appears again.
maybe generate in 720p. it all fixed for me after doing that
Anyone else has their video slightly lose color contrast over time? Been trying to fix this for a few days and I am not sure what's doing it.
nice
Most of the workflows I download never work. Comfyui updates disable or don't allow node
Nope, this workflow is incomplete.