6 step process with number of frames > 81: 2 steps high noise with context options, 1 step high noise w/o context options, 3 steps low noise without context options
This trick is good. Use Lightning Lora high and low at strength 1 + Lora Lightning WAN2.1 at strength 3 on the high, and 0.25 on the low You can do 4 steps, with 2 samplers, 2 + 2, good motion, no more slow motion.
Resized original image to 512x512, made WAN 2.2 video, then upscaled it via FlashVSR to 1024x1024 - good result on a face