The result has some problem

[Vickeyhw]

I use the wan 2.2 version, with the default command to infer. But the generated video has some color problem except the first frame.

liuyifei.mp4

[picolo100]

I use the wan 2.2 version, with the default command to infer. But the generated video has some color problem except the first frame.

liuyifei.mp4

lightx2v issue 100% or settings in sampler

[Vickeyhw]

@picolo100 Do you have a solution? I am using the default inference script without modifying any configurations or loading the acceleration LoRA provided by lightx2v.

[picolo100]

@picolo100 Do you have a solution? I am using the default inference script without modifying any configurations or loading the acceleration LoRA provided by lightx2v.

could you bring your workflow .json?

[Vickeyhw]

@picolo100 I did not use workflow. just execute the python script by
python infer.py \ --ip_image "/opt/nas/p/data/people/5.jpg" \ --prompt "A woman sits comfortably at a sofa" \ --output "/test/output/liuyifei.mp4" \ --wan_version '2.2' --fps 16

[KBRASK]

Hello! Thanks for your interest in Stand-In. Based on my testing so far, I haven’t encountered this kind of issue. Could you share the relevant inputs so I can try to reproduce the problem?

[Vickeyhw]

@KBRASK I use this command to infer:
python infer.py \ --ip_image "people/5.jpg" \ --prompt "A woman sits comfortably at a sofa" \ --output "/test/output/liuyifei.mp4" \ --wan_version '2.2' --fps 16 with this image:

[picolo100]

@picolo100 I did not use workflow. just execute the python script by python infer.py \ --ip_image "/opt/nas/p/data/people/5.jpg" \ --prompt "A woman sits comfortably at a sofa" \ --output "/test/output/liuyifei.mp4" \ --wan_version '2.2' --fps 16

i though you was using it throgh comfyui, try --fp32

[picolo100]

Hello! Thanks for your interest in Stand-In. Based on my testing so far, I haven’t encountered this kind of issue. Could you share the relevant inputs so I can try to reproduce the problem?

could you help us in comfyui? i guess it's more convenient nowadays to work in comfy to have more manual handling stuff to optimize the quality output, im here to seeking facial consistency toward the animation in my wan animate workflow, do you believe you could help me on this, im using Lynx for now, but if the recent stand in update does better i would definitely go with stand -in. so lemme know