I'd like to share my workhorse workflow for Wan2.2 video generation. Since many people complained about Lighting LoRA messing with output video dynamics - I decided to include an additional step that launches a high-noise model without any LoRA before passing the latent video to the main "high_noise + low_noise" sampler.
I don't usually enable frame interpolation and upscaling but there are 2 optional steps in this workflow for that.
Required ComfyUI custom nodes for the workflow: