This is a spin on the original Kijai's Wan 2.2 Animate Workflow to make it more accessible to low VRAM GPU Cards.

⚠ If in doubt or OOM errors, read the comments inside the yellow boxes in the workflow ⚠
⚠ This WF uses SageAttention. If you don't have or need Triton/Torch/SageAttention you'll have to disconnect the "WanVideo Torch Compile Settings" in the "Model Loader" group or else you'll have a Triton related error. ⚠
❕❕ Tested with 12GB VRAM / 32GB RAM
❕❕ I was able to generate 113 Frames @ 640p with this setup (10min)
❕❕ Use the Download button at the top right in this page
🟣 All important nodes are colored Purple

Main differences:

• VAE precision set to fp16 instead of fp32

• FP8 Scaled Text Encoder instead of FP16
  (If you prefer the FP16 just copy from the Kijai's original wf node and replace my prompt setup)

• Video and Image resolutions are calculated automatically

• Fast Enable/Disable functions (Masking, Face Tracking, etc.)

• Easy Frame Window Size setting

I tried to organize everything without hiding anything, this way it should be better for newcomers to understand the workflow process.